**Product Manager, Credit Card**\n\nWe’re looking for an entrepreneurial product manager to take a lead role in continuing to build the first health and wellness credit card

Limit

limit

London, UK, April 2nd, 2025/Chainwire/--Decentraland's flagship virtual event returns April 9–12, showcasing digital fashion as a new canvas for storytelling, activism, and global identity. Metaverse Fashion Week (MVFW) returns to Decentraland from April 9–12, amid a shifting fashion landscape characterized by declining traditional luxury sales and evolving consumer preferences. Metaverse Fashion Week Metaverse Fashion Week MVFW 2025 highlights digital fashion's growing relevance as a meaningful response to these industry transformations. Under this year's theme, "Infinite Identities," more than 47 global designers and brands leverage digital mediums to challenge creative boundaries and articulate nuanced cultural identities. Hosted on Decentraland 2.0, the platform’s newly redesigned desktop client offers improved visuals, greater interactivity, and better accessibility. These enhancements, developed during the event's hiatus, provide an immersive and engaging environment for attendees and creators alike. Decentraland 2.0 Decentraland 2.0 “This year’s Metaverse Fashion Week comes at a critical moment for fashion and our society at large; one marked by rethinking priorities, authenticity, and connection," says Bay Backner, Head Producer of MVFW. “This year’s Metaverse Fashion Week comes at a critical moment for fashion and our society at large; one marked by rethinking priorities, authenticity, and connection," says Bay Backner, Head Producer of MVFW. “I’ve had the privilege of working closely with every designer and brand taking part, and I’ve seen how digital fashion can elevate unheard voices, ask important questions, and spark real cultural change. In a time like this, when the physical world feels increasingly heavy or limiting, third spaces like Decentraland become places where people can breathe, create, and truly be themselves. MVFW 2025 channels that energy into a live moment for fashion’s next breakout names and creative economies.” “I’ve had the privilege of working closely with every designer and brand taking part, and I’ve seen how digital fashion can elevate unheard voices, ask important questions, and spark real cultural change. In a time like this, when the physical world feels increasingly heavy or limiting, third spaces like Decentraland become places where people can breathe, create, and truly be themselves. MVFW 2025 channels that energy into a live moment for fashion’s next breakout names and creative economies.” Runways: Fashion as Cultural Dialogue Runways have always been the cornerstone of fashion weeks worldwide, but MVFW 2025 reimagines this classic format through immersive, collaborative, and interactive experiences. Attendees can anticipate creative surprises as designers redefine the runway with innovative animated collections that blur the line between digital and physical fashion. Cyberdog: London’s iconic brand brings its neon-infused, rave-culture aesthetic into the virtual realm, blending futuristic fashion with underground electronic subcultures. Free The Youth: Ghanaian collective Free The Youth highlights African streetwear and creativity, celebrating bold, culturally driven designs that bridge physical and virtual worlds. The Banners We Wear: Decentraland’s inaugural MVFW Prize unveils deeply personal collections from emerging designers. With a $15,000 prize pool, this unique competition challenged designers globally to create digital wearables that convey powerful messages about identity, activism, and resilience. Winning entries will debut on a dedicated runway, spotlighting urgent social narratives and elevating voices often marginalized in mainstream fashion. Interactive Designer Showrooms: Digital Fashion as Storytelling At the heart of MVFW 2025 are 16 interactive designer showrooms, each exploring the interplay between fashion, technology, and cultural narrative. These showrooms are divided into two thematic zones, Augmented Self and Storied Self, offering distinct yet complementary perspectives on digital self-expression. Augmented Self: This zone highlights innovative uses of AI, AR, and digital-physical ("phygital") fashion. Showrooms include AI-driven styling by Digital Drip, AR-enhanced virtual fashion by FFFACE.ME, NFT-powered wearable collaborations from Another-1 X Templa, and microchip manicure technology from Chipped. Augmented Self: This zone highlights innovative uses of AI, AR, and digital-physical ("phygital") fashion. Showrooms include AI-driven styling by Digital Drip, AR-enhanced virtual fashion by FFFACE.ME, NFT-powered wearable collaborations from Another-1 X Templa, and microchip manicure technology from Chipped. Storied Self: Dedicated to fashion as cultural storytelling, this zone spotlights global narratives and identity-driven collections. Key highlights include adaptive fashion for prosthetics by Ukraine's SVD Denim, mycelium-inspired digital wearables from Amsterdam's GLTCH Atelier, and culturally resonant digital storytelling from Puerto Rico’s Spatialgineers Collective. Storied Self: Dedicated to fashion as cultural storytelling, this zone spotlights global narratives and identity-driven collections. Key highlights include adaptive fashion for prosthetics by Ukraine's SVD Denim, mycelium-inspired digital wearables from Amsterdam's GLTCH Atelier, and culturally resonant digital storytelling from Puerto Rico’s Spatialgineers Collective. Immersive Experiences & Insightful Conversations Beyond runways and showrooms, MVFW 2025 invites attendees to experience transformative fashion installations and conversations exploring identity, technology, and culture: "Entangled Q" by Cansy: A quantum-inspired art installation featuring avatars that dynamically morph, exploring fluid identities. "Entangled Q" by Cansy: A quantum-inspired art installation featuring avatars that dynamically morph, exploring fluid identities. "OFF-WORLD Galactic Expansion" by The Immersive KIND: A futuristic display merging AI-generated beauty, robotics couture, and interstellar aesthetics. "OFF-WORLD Galactic Expansion" by The Immersive KIND: A futuristic display merging AI-generated beauty, robotics couture, and interstellar aesthetics. A curated series of panels will dive deeper into key conversations, covering the evolving role of avatars in digital self-expression, AI’s impact on fashion creativity, wearable tech in gaming environments, and the intersection of physical and virtual retail. Panelists include industry leaders from Harper’s BAZAAR Australia, Parsons School of Design, Ready Player Me, Institut Français de la Mode, Boy’s Club, and more. Metaverse Fashion Week 2025 taps into a growing appetite for fashion that has meaning beyond trends, signaling a shift toward stories that matter and identities that resonate globally. As digital spaces become essential platforms for expression, Decentraland opens new possibilities for creators and communities to craft a freer, more inclusive vision of fashion. With immersive runways, interactive storytelling, and conversations that challenge the status quo, MVFW 2025 invites everyone to experience how digital fashion can reflect—and reshape—our world. To learn more and get ready for MVFW 2025, users can visit mvfw.org and download the latest client. mvfw.org mvfw.org About Decentraland: Decentraland is the first decentralized, community-driven social virtual world. It empowers users to create, explore, and connect within an open, immersive digital landscape where they fully own their digital assets. Decentraland Decentraland Whether hosting virtual events or designing their digital identity, individuals can shape a world that prioritizes self-expression and freedom—free from the constraints of corporate control. Governed by its community through a decentralized autonomous organization (DAO) and supported by a non-profit Foundation, Decentraland operates as an open-source, traversable world that encourages creativity, ownership, and innovation. By fostering collaboration and community, Decentraland is shaping the future of digital interaction, where individuals can truly own and build their digital lives. Contact Ana Lezama press@decentraland.org This story was distributed as a release by Chainswire under HackerNoon’s Business Blogging Program. Learn more about the program here This story was distributed as a release by Chainswire under HackerNoon’s Business Blogging Program. Learn more about the program here here here

Global Designers Challenge Borders, Identity, And The Limits Of Physical Fashion At MVFW 2025

Qlobal Dizaynerlər MFW 2025-də Sərhədlərə, Şəxsiyyətə və Fiziki Moda Sərhədlərinə meydan oxuyurlar

Globalni dizajneri izazivaju granice, identitet i granice fizičke mode na MFW 2025.

Diseñadores globales desafían las fronteras, la identidad y los límites de la moda física en MFW 2025

ERROR: invalid target_lang (iw)

グローバルデザイナーがMFW2025で物理ファッションの境界、アイデンティティ、そして限界に挑戦

អ្នករចនាសម្ព័ន្ធទូទាំងពិភពលោកជំរុញកម្រិតខុសគ្នានិងកម្រិតខុសគ្នានៃភាពរចនាសម្ព័ន្ធរចនាសម្ព័ន្ធនៅ MFW 2025 ។

Globālie dizaineri izaicina robežas, identitāti un fiziskās modes robežas MFW 2025

ERROR: invalid target_lang (nso)

Designerii globali provoacă granițele, identitatea și limitele modei fizice la MFW 2025

ගෝලීය නිර්මාණකරුවන් MFW 2025 හි භෞතික විලාසිතාගේ සීමාවන්, අනන් යතාව සහ සීමාවන් අභියෝග කරයි

Global Vagadziri Vanodenha Border, Identity, Uye Iyo Miganhu YePhysical Fashoni PaMFW 2025

Global designers challenge border, identity, and the limits of physical fashion at MFW 2025. - Global designers challenge borders, identity, and the limits of physical fashion at MFW 2025

Table of Links Abstract and 1. Introduction Abstract and 1. Introduction 2 Data 2.1 Data Sources 2.1 Data Sources 2.2 SS and SI Categories 2.2 SS and SI Categories 3 Methods 3.1 Lexicon Creation and Expansion 3.1 Lexicon Creation and Expansion 3.2 Annotations 3.2 Annotations 3.3 System Description 3.3 System Description 4 Results 4.1 Demographics and 4.2 System Performance 4.1 Demographics and 4.2 System Performance 5 Discussion 5 Discussion 5.1 Limitations 5.1 Limitations 6 Conclusion, Reproducibility, Funding, Acknowledgments, Author Contributions, and References 6 Conclusion, Reproducibility, Funding, Acknowledgments, Author Contributions, and References SUPPLEMENTARY SUPPLEMENTARY Guidelines for Annotating Social Support and Social Isolation in Clinical Notes Guidelines for Annotating Social Support and Social Isolation in Clinical Notes Other Supervised Models Other Supervised Models 5.1 Limitations Several limitations should be noted. There were insufficient instances in the notes of the emotional support subcategories to evaluate the NLP systems. Emotional support (and lack thereof) is an important and distinct fine-grained category that would ideally be identified in the notes. Second, the RBS was designed with specific lexicons from manual review at MSHS and WCM, may have experienced overfitting and led to an inflated f-score. It would be beneficial to validate these NLP systems on clinical notes from different EHR systems. Other healthcare systems that implement a lexicon-based rules approach will need to perform site-specific template removal to avoid the problem of false-positives. With fine-tuning, the LLM approach may have been able to correctly interpret the templates; however, because the templates were removed from the notes before the annotation process, this was not assessed. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv Authors:
(1) Braja Gopal Patra, Weill Cornell Medicine, New York, NY, USA and co-first authors;
(2) Lauren A. Lepow, Icahn School of Medicine at Mount Sinai, New York, NY, USA and co-first authors;
(3) Praneet Kasi Reddy Jagadeesh Kumar. Weill Cornell Medicine, New York, NY, USA;
(4) Veer Vekaria, Weill Cornell Medicine, New York, NY, USA;
(5) Mohit Manoj Sharma, Weill Cornell Medicine, New York, NY, USA;
(6) Prakash Adekkanattu, Weill Cornell Medicine, New York, NY, USA;
(7) Brian Fennessy, Icahn School of Medicine at Mount Sinai, New York, NY, USA;
(8) Gavin Hynes, Icahn School of Medicine at Mount Sinai, New York, NY, USA;
(9) Isotta Landi, Icahn School of Medicine at Mount Sinai, New York, NY, USA;
(10) Jorge A. Sanchez-Ruiz, Mayo Clinic, Rochester, MN, USA;
(11) Euijung Ryu, Mayo Clinic, Rochester, MN, USA;
(12) Joanna M. Biernacka, Mayo Clinic, Rochester, MN, USA;
(13) Girish N. Nadkarni, Icahn School of Medicine at Mount Sinai, New York, NY, USA;
(14) Ardesheer Talati, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, USA and New York State Psychiatric Institute, New York, NY, USA;
(15) Myrna Weissman, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, USA and New York State Psychiatric Institute, New York, NY, USA;
(16) Mark Olfson, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, USA, New York State Psychiatric Institute, New York, NY, USA, and Columbia University Irving Medical Center, New York, NY, USA;
(17) J. John Mann, Columbia University Irving Medical Center, New York, NY, USA;
(18) Alexander W. Charney, Icahn School of Medicine at Mount Sinai, New York, NY, USA;
(19) Jyotishman Pathak, Weill Cornell Medicine, New York, NY, USA. Authors: Authors: (1) Braja Gopal Patra, Weill Cornell Medicine, New York, NY, USA and co-first authors; (2) Lauren A. Lepow, Icahn School of Medicine at Mount Sinai, New York, NY, USA and co-first authors; (3) Praneet Kasi Reddy Jagadeesh Kumar. Weill Cornell Medicine, New York, NY, USA; (4) Veer Vekaria, Weill Cornell Medicine, New York, NY, USA; (5) Mohit Manoj Sharma, Weill Cornell Medicine, New York, NY, USA; (6) Prakash Adekkanattu, Weill Cornell Medicine, New York, NY, USA; (7) Brian Fennessy, Icahn School of Medicine at Mount Sinai, New York, NY, USA; (8) Gavin Hynes, Icahn School of Medicine at Mount Sinai, New York, NY, USA; (9) Isotta Landi, Icahn School of Medicine at Mount Sinai, New York, NY, USA; (10) Jorge A. Sanchez-Ruiz, Mayo Clinic, Rochester, MN, USA; (11) Euijung Ryu, Mayo Clinic, Rochester, MN, USA; (12) Joanna M. Biernacka, Mayo Clinic, Rochester, MN, USA; (13) Girish N. Nadkarni, Icahn School of Medicine at Mount Sinai, New York, NY, USA; (14) Ardesheer Talati, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, USA and New York State Psychiatric Institute, New York, NY, USA; (15) Myrna Weissman, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, USA and New York State Psychiatric Institute, New York, NY, USA; (16) Mark Olfson, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, USA, New York State Psychiatric Institute, New York, NY, USA, and Columbia University Irving Medical Center, New York, NY, USA; (17) J. John Mann, Columbia University Irving Medical Center, New York, NY, USA; (18) Alexander W. Charney, Icahn School of Medicine at Mount Sinai, New York, NY, USA; (19) Jyotishman Pathak, Weill Cornell Medicine, New York, NY, USA.

NLP Performance in Clinical Notes: Addressing Data Limitations and System Overfitting

Willemstad, Curaçao, Netherlands, March 31st, 2025/GamingWire/--Cloudbet is opening no-limit betting for April's Dota tournament blitz, seizing the moment as the start of the year shapes up to be Dota 2's busiest season outside of The International. With ESL One Raleigh 2025 (April 7-13) and PGL Wallachia Season 4 (April 18-26) both offering million-dollar prize pools on the heels of the recent 7.38c patch, competitive Dota is about to get interesting. This means serious Dota bettors on Cloudbet can now tap into the combined resources of global markets, purpose-built for high-stakes action – meaning serious bettors could potentially win more than the tournaments' champions themselves. Dota bettors on Cloudbet Dota bettors on Cloudbet Cloudbet's no-limit bets now cover five essential Dota markets: Dota markets Dota markets Map Handicap – Allows wagers on matches where underdogs are given an advantage.
Total Maps – Enables bets on the total number of maps played in a series.
Winner – Offers the option to bet on the overall winner of the series.
Map Winner – Provides betting on the outcome of individual maps.
Correct Map Score – Allows predictions of the final map score for a series. Map Handicap – Allows wagers on matches where underdogs are given an advantage. Total Maps – Enables bets on the total number of maps played in a series. Winner – Offers the option to bet on the overall winner of the series. Map Winner – Provides betting on the outcome of individual maps. Correct Map Score – Allows predictions of the final map score for a series. The timing is particularly notable with the 7.38c patch update that dropped during the FISSURE Universe playoffs — shaking up hero rankings and team strategies right in the middle of high-stakes competitions. "We're listening to all the feedback in our esports community about how to bring them closer to games all season long." notes John, the Cloudbet Spokesperson. "And we're doing that with no-limit bets, and integrating many, many more markets. You can bet on which team gets first blood, first barracks, or first Roshan – markets that go beyond the basics but are core to the Dota experience. These are milestone events that make grown adults scream at their monitors, so why not cash in too?” "We're listening to all the feedback in our esports community about how to bring them closer to games all season long." notes John, the Cloudbet Spokesperson. "And we're doing that with no-limit bets, and integrating many, many more markets. You can bet on which team gets first blood, first barracks, or first Roshan – markets that go beyond the basics but are core to the Dota experience. These are milestone events that make grown adults scream at their monitors, so why not cash in too?” New Cloudbet users will also have access to the Welcome Package, featuring up to $2500 in cash, ongoing Rakeback Rewards, and straight cash rewards through the Vault. Once claimed, users enter Cloudbet Rewards, the generous loyalty program designed for consistent bettors. While traditional sportsbooks often limit the size of bets, especially in esports, Cloudbet offering no-limit betting on major Dota 2 markets is an industry first. In high-integrity traditional sports like the NFL and English Premier League, betting limits can reach into the millions. By extending its no-limit betting, already available on major sports, to Dota 2, Cloudbet aims to bring esports into the top tier of the sports betting world. For Dota fans looking to add another dimension to their viewing experience, Cloudbet's no-limit betting offers a way to back predictions with real conviction – and potentially serious returns. For more information, users can visit Cloudbet.com. Cloudbet.com Cloudbet.com ABOUT CLOUDBET Founded in 2013, Cloudbet is the world’s longest-running crypto casino and sportsbook. Over the past decade, players worldwide have placed millions of bets using over 40 different cryptocurrencies. In 2024, Cloudbet introduced the most generous welcome offer and loyalty program online, featuring stacked rewards and guaranteed daily cash drops for frequent bettors. Cloudbet Cloudbet With a wide selection of slots, live casino games, and sports markets—ranging from esports to Premier League and NFL player props—Cloudbet is the leader in secure crypto betting. Users can find more information at Cloudbet.com; Instagram (@cloudbetofficial); Twitter/X (@Cloudbet). Cloudbet.com Cloudbet.com Instagram (@cloudbetofficial) Instagram (@cloudbetofficial) Twitter/X (@Cloudbet) Twitter/X (@Cloudbet) Contact Irene Halcyon Super Holdings B.V. irene@ This story was distributed as a release by Gamingwire under HackerNoon’s Business Blogging Program. Learn more about the program here This story was distributed as a release by Gamingwire under HackerNoon’s Business Blogging Program. Learn more about the program here here here

Cloudbet Opens No-limit Betting For April's Dota Tournament Blitz

Cloudbet يفتتح عرض غير محدود لبرنامج Dota Tournament Blitz في أبريل

Cloudbet отваря безлимитни залози за Dota Tournament Blitz през април

Cloudbet eröffnet No-Limit-Wetten für April-Dota-Turnier Blitz

Cloudbet abre apuestas sin límite para el torneo Blitz de Dota de abril

Cloudbet avaa No-limit-vedonlyönnin huhtikuun Dota Tournament Blitzille

Cloudbet、4月のDotaトーナメントブリッツでノーリミットベッティングを開始

Cloudbet открывает безлимитные ставки на апрельский турнир Dota Blitz

අඟහරුවාදා Dota Tournament Blitz සඳහා No-Limit Betting විවෘත කරයි

Cloudbet hap bastet pa kufi për Blitz të Turneut Dota të prillit

ክላውድቤት ንናይ ሚያዝያ ዶታ ውድድር ብሊትዝ ደረት ኣልቦ ውርርድ ከፊቱ

Cloudbet, Nisan Dota Turnuvası Blitz'i İçin Limitsiz Bahis Açıyor

Welcome back to another HackerNoon Startups of the Week! If you’re not already in the loop, each week, the team at HackerNoon showcases a list of startups from our Startups of The Year database. All these startups have been nominated as one of the best in their respective category or region. This week, we’re pleased to bring you FoundersToday, The Forbes Group, and Bloklab. Want to be nominated for HackerNoon’s Startups of The Year? Learn how here. Meet Startups of the Week FoundersToday FoundersToday is a Startup Publication Media that delivers the latest news, insights, and trends in the startup ecosystem. It offers high-quality, unbiased, and thoroughly researched content that covers a wide range of topics, including emerging technologies, venture capital, startup strategies, and entrepreneurial journeys. Located in Linz, Austria, this startup has been nominated as one of the best in their region, along with nominations in the News & Journalism, Publishing, and Writing and Editing categories. Support FoundersToday - Vote here! The Forbes Group The Forbes Group is a full-service consultancy based in Calgary, Alberta. It specializes in aligning businesses with modern value chain strategies that drive sustainability compliance, operational excellence, and measurable growth. Located in Calgary, Canada, this startup has been nominated as one of the best in their region, along with nominations in the Leadership, Professional Services, and Training and Consulting categories. Support The Forbes Group - Vote here! Bloklab Bloklab Oy is Finland's leading technology start-up in blockchain technologies and decentralized finance (DeFi). With a team of experts, Bloklab Oy develops innovative platforms that empower and optimize the digital assets of businesses, adhering to principles of security, transparency, and user-centricity. Located in Helsinki, Finland, this startup has been nominated as one of the best in their region, along with nominations in the Blockchain and Decentralization categories. Support Bloklab - Vote here! Featured Interview of The Week Follow the example of FoundersToday and create your business page on HackerNoon. Having a business profile will allow you to create your very own Evergreen Tech Company News Page and participate in our Tech Companies ranking, weekly and data-driven insights that delve into which tech companies are gaining and losing traction in public awareness. Plus, only with a business page can you answer the awesome interview templates we crafted for you. Today, we’re focusing on the Sao Paulo Startup Interview Template. Sao Paulo was one of our trending startup cities last year, and this edition is no different! If you’re a Sao Paulo-based start up or have insights to share about the tech scene in one of Latin America’s largest cities, this one’s for you! Use this template to showcase your startup’s vision, innovative projects, key lessons, and strategies for overcoming challenges in the city. Insights about how the city evolves and supports entrepreneurs are welcome! Additionally, don’t forget to share how events like HackerNoon’s Startups of the Year have shaped your journey. Highlight why your startup stands out and deserves recognition in 2024. Here’s a quote from FoundersToday featured in their published interview, telling us why they decided to participate in Startups of The Year: We’re excited to participate in HackerNoon’s Startups of the Year because it provides a fantastic platform to share our journey at FoundersToday with a broader tech-savvy audience. HackerNoon’s community is deeply embedded in the startup and entrepreneurial ecosystem, making it an ideal space to highlight our vision, challenges, and successes. By participating, we aim to connect with like-minded entrepreneurs, investors, and tech enthusiasts who can offer valuable insights and collaborations to help us grow further. Want to be featured on HackerNoon’s Startups of the Week? Share your startup's story - use this interview template. Here’s how the 2023 nominees did it: Meetanshi, Sniper.xyz, Wallet Factory, Grownu Workforce Management System, Motif, Lightrun Featured Startups Special Package Discover our Lead Generation package, delivering 1M impressions through the HackerNoon Newsletter! With this package, you’ll get: 4 Total Newsletters Sent: 325,000 per send (85% of subscribers reside in the US)

Open Rate: 13% - 15%
Click-to-Open Rate: 13.61%
Sample: Here


Your Evergreen Tech Company News Page Learn more about this package here or book a meeting with us! About HackerNoon’s Startups of The Year Startups of The Year 2024 is HackerNoon’s flagship community-driven event celebrating startups, technology, and the spirit of innovation. Currently in its third iteration, the prestigious Internet award recognizes and celebrates tech startups of all shapes and sizes. This year, over 150,000 entities across 4200+ cities, 6 continents, and 100+ industries will participate in a bid to be crowned the best startup of the year! Millions of votes have been cast over the past few years, and many stories have been written about these daring and rising startups. The winners will get a free interview on HackerNoon and an Evergreen Tech Company News page. Visit our FAQ page to learn more. Download our design assets here. Check out the Startups of the Year Merch Shophere. HackerNoon’s Startups of The Year is a branding opportunity unlike any other. Whether your goal is brand awareness or lead generation, HackerNoon has curated startup-friendly packages to solve your marketing challenges. Meet our sponsors: Wellfound: Join the #1 global, startup-focused community. At Wellfound, we're not just a job board—we're the place where top startup talent and the world's most exciting companies connect to build the future. Notion: Notion is trusted and loved by thousands of startups as their connected workspace—from building product roadmaps to tracking fundraising. Try Notion with unlimited AI, FREE for up to 6 months, to build and scale your company with one powerful tool. Get your offer now! Hubspot: If you’re looking for a smart CRM platform that meets the needs of small businesses, look no further than HubSpot. Seamlessly connect your data, teams and customers in one easy-to-use scalable platform that grows with your business. Get started for free. Bright Data: Startups that leverage public web data can make faster, data-driven decisions, giving them a competitive edge. With Bright Data’s scalable web data collection, businesses can grow from a small operation to an enterprise by harnessing insights at every stage. Algolia: Algolia NeuralSearch is the world's only AI end-to-end Search and Discovery Platform combining powerful keyword and natural language processing in a single API. Welcome back to another HackerNoon Startups of the Week! If you’re not already in the loop, each week, the team at HackerNoon showcases a list of startups from our Startups of The Year database . All these startups have been nominated as one of the best in their respective category or region. Startups of The Year database This week, we’re pleased to bring you FoundersToday , The Forbes Group , and Bloklab . FoundersToday The Forbes Group Bloklab Want to be nominated for HackerNoon’s Startups of The Year? Learn how here. Want to be nominated for HackerNoon’s Startups of The Year? Learn how here. Want to be nominated for HackerNoon’s Startups of The Year? Learn how here . here Meet Startups of the Week Meet Startups of the Week FoundersToday FoundersToday FoundersToday FoundersToday is a Startup Publication Media that delivers the latest news, insights, and trends in the startup ecosystem. It offers high-quality, unbiased, and thoroughly researched content that covers a wide range of topics, including emerging technologies, venture capital, startup strategies, and entrepreneurial journeys. Located in Linz, Austria , this startup has been nominated as one of the best in their region, along with nominations in the News & Journalism , Publishing , and Writing and Editing categories. Linz, Austria News & Journalism Publishing Writing and Editing Support FoundersToday - Vote here! Support FoundersToday - Vote here! Support FoundersToday - Vote here ! Vote here The Forbes Group The Forbes Group The Forbes Group The Forbes Group is a full-service consultancy based in Calgary, Alberta. It specializes in aligning businesses with modern value chain strategies that drive sustainability compliance, operational excellence, and measurable growth. Located in Calgary, Canada , this startup has been nominated as one of the best in their region, along with nominations in the Leadership , Professional Services , and Training and Consulting categories. Calgary, Canada Leadership Professional Services Training and Consulting Support The Forbes Group - Vote here! Support The Forbes Group - Vote here! Support The Forbes Group - Vote here ! Vote here Bloklab Bloklab Bloklab Bloklab Oy is Finland's leading technology start-up in blockchain technologies and decentralized finance (DeFi). With a team of experts, Bloklab Oy develops innovative platforms that empower and optimize the digital assets of businesses, adhering to principles of security, transparency, and user-centricity. Located in Helsinki, Finland , this startup has been nominated as one of the best in their region, along with nominations in the Blockchain and Decentralization categories. Helsinki, Finland Blockchain Decentralization Support Bloklab - Vote here! Support Bloklab - Vote here! Support Bloklab - Vote here ! Vote here Featured Interview of The Week Featured Interview of The Week Follow the example of FoundersToday and create your business page on HackerNoon . FoundersToday your business page on HackerNoon Having a business profile will allow you to create your very own Evergreen Tech Company News Page and participate in our Tech Companies ranking , weekly and data-driven insights that delve into which tech companies are gaining and losing traction in public awareness. Plus, only with a business page can you answer the awesome interview templates we crafted for you. Evergreen Tech Company News Page Tech Companies ranking Today, we’re focusing on the Sao Paulo Startup Interview Template . Sao Paulo was one of our trending startup cities last year, and this edition is no different! If you’re a Sao Paulo-based start up or have insights to share about the tech scene in one of Latin America’s largest cities, this one’s for you! Use this template to showcase your startup’s vision, innovative projects, key lessons, and strategies for overcoming challenges in the city. Insights about how the city evolves and supports entrepreneurs are welcome! Additionally, don’t forget to share how events like HackerNoon’s Startups of the Year have shaped your journey. Highlight why your startup stands out and deserves recognition in 2024. Sao Paulo Startup Interview Template trending startup cities Use this template Here’s a quote from FoundersToday featured in their published interview , telling us why they decided to participate in Startups of The Year: FoundersToday published interview We’re excited to participate in HackerNoon’s Startups of the Year because it provides a fantastic platform to share our journey at FoundersToday with a broader tech-savvy audience. HackerNoon’s community is deeply embedded in the startup and entrepreneurial ecosystem, making it an ideal space to highlight our vision, challenges, and successes. By participating, we aim to connect with like-minded entrepreneurs, investors, and tech enthusiasts who can offer valuable insights and collaborations to help us grow further. We’re excited to participate in HackerNoon’s Startups of the Year because it provides a fantastic platform to share our journey at FoundersToday with a broader tech-savvy audience. HackerNoon’s community is deeply embedded in the startup and entrepreneurial ecosystem, making it an ideal space to highlight our vision, challenges, and successes. By participating, we aim to connect with like-minded entrepreneurs, investors, and tech enthusiasts who can offer valuable insights and collaborations to help us grow further. Want to be featured on HackerNoon’s Startups of the Week? Share your startup's story - use this interview template. Here’s how the 2023 nominees did it: Meetanshi, Sniper.xyz, Wallet Factory, Grownu Workforce Management System, Motif, Lightrun Want to be featured on HackerNoon’s Startups of the Week? Share your startup's story - use this interview template. Here’s how the 2023 nominees did it: Meetanshi , Sniper.xyz , Wallet Factory , Grownu Workforce Management System , Motif , Lightrun Want to be featured on HackerNoon’s Startups of the Week? Share your startup's story - use this interview template . use this interview template Meetanshi Meetanshi Sniper.xyz Sniper.xyz Wallet Factory Wallet Factory Grownu Workforce Management System Grownu Workforce Management System Motif Motif Lightrun Lightrun Featured Startups Special Package Discover our Lead Generation package, delivering 1M impressions through the HackerNoon Newsletter! Lead Generation With this package, you’ll get: With this package, you’ll get: 4 Total Newsletters Sent: 325,000 per send (85% of subscribers reside in the US)

Open Rate: 13% - 15%
Click-to-Open Rate: 13.61%
Sample: Here Your Evergreen Tech Company News Page 4 Total Newsletters Sent: 325,000 per send (85% of subscribers reside in the US) Open Rate: 13% - 15%
Click-to-Open Rate: 13.61%
Sample: Here Open Rate: 13% - 15% Click-to-Open Rate: 13.61% Sample: Here Open Rate: 13% - 15% Click-to-Open Rate: 13.61% Sample: Here Here Your Evergreen Tech Company News Page Evergreen Tech Company News Page Learn more about this package here or book a meeting with us! Learn more about this package here or book a meeting with us ! here here book a meeting with us book a meeting with us About HackerNoon’s Startups of The Year Startups of The Year 2024 is HackerNoon’s flagship community-driven event celebrating startups, technology, and the spirit of innovation. Currently in its third iteration, the prestigious Internet award recognizes and celebrates tech startups of all shapes and sizes. This year, over 150,000 entities across 4200+ cities, 6 continents, and 100+ industries will participate in a bid to be crowned the best startup of the year! Millions of votes have been cast over the past few years, and many stories have been written about these daring and rising startups. Startups of The Year 2024 many stories The winners will get a free interview on HackerNoon and an Evergreen Tech Company News page. free interview Evergreen Tech Company News Visit our FAQ page to learn more. FAQ Download our design assets here . here Check out the Startups of the Year Merch Shop here . here HackerNoon’s Startups of The Year is a branding opportunity unlike any other. Whether your goal is brand awareness or lead generation, HackerNoon has curated startup-friendly packages to solve your marketing challenges. startup-friendly packages Meet our sponsors: Meet our sponsors: Wellfound: Join the #1 global, startup-focused community . At Wellfound, we're not just a job board—we're the place where top startup talent and the world's most exciting companies connect to build the future. Wellfound: Join the #1 global, startup-focused community Notion: Notion is trusted and loved by thousands of startups as their connected workspace—from building product roadmaps to tracking fundraising. Try Notion with unlimited AI, FREE for up to 6 months , to build and scale your company with one powerful tool. Get your offer now ! Notion: Try Notion with unlimited AI, FREE for up to 6 months Get your offer now ! Hubspot: If you’re looking for a smart CRM platform that meets the needs of small businesses, look no further than HubSpot. Seamlessly connect your data, teams and customers in one easy-to-use scalable platform that grows with your business. Get started for free . Hubspot: Get started for free Bright Data: Startups that leverage public web data can make faster, data-driven decisions, giving them a competitive edge. With Bright Data’s scalable web data collection , businesses can grow from a small operation to an enterprise by harnessing insights at every stage. Bright Data: Bright Data’s scalable web data collection Algolia: Algolia NeuralSearch is the world's only AI end-to-end Search and Discovery Platform combining powerful keyword and natural language processing in a single API. Algolia: AI end-to-end Search and Discovery Platform

Meet FoundersToday, Forbes Group, Bloklab Limited: HackerNoon Startups of the Week

Conozca a FoundersToday, Forbes Group y Bloklab Limited: Startups de la semana de HackerNoon

フォーブス・グループ、Blocklab Limited: HackerNoon Startups of the Week

Mayi - Mayi - Mayi - Mayi - Mayi - Mayi

Susipažinkite su „Forbes Group“ įkūrėjais, „Bloklab Limited“: „HackerNoon“ savaitės startuoliai

Forbes Group, Bloklab Limited, үүсгэн байгуулагчидтай өнөөдөр уулзаарай: Долоо хоногийн шилдэг стартапууд.

Kamachiqkunawan riqsinakuyKunan punchaw, Forbes Grupo, Bloklab Limited: HackerNoon Simana qallariyninkuna

Shirkadda Forbes Group, Bloklab Limited: HackerNoon Startups of the Week

Упознајте ФоундерсТодаи, Форбес Гроуп, Блоклаб Лимитед: ХацкерНоон Стартупс оф тхе Веек

ERROR: invalid target_lang (tk)

Forumlar, Forbes Group, Bloklab Limited: HackerNoon Startups of the Week – HackerNoon Group: Forbes Group, Bloklab Limited: HackerNoon Startups of the Week

Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Works 2 Related Works 3 Preliminaries 3.1 Fair Supervised Learning and 3.2 Fairness Criteria 3.1 Fair Supervised Learning and 3.2 Fairness Criteria 3.3 Dependence Measures for Fair Supervised Learning 3.3 Dependence Measures for Fair Supervised Learning 4 Inductive Biases of DP-based Fair Supervised Learning 4 Inductive Biases of DP-based Fair Supervised Learning 4.1 Extending the Theoretical Results to Randomized Prediction Rule 4.1 Extending the Theoretical Results to Randomized Prediction Rule 5 A Distributionally Robust Optimization Approach to DP-based Fair Learning 5 A Distributionally Robust Optimization Approach to DP-based Fair Learning 6 Numerical Results 6.1 Experimental Setup 6.1 Experimental Setup 6.2 Inductive Biases of Models trained in DP-based Fair Learning 6.2 Inductive Biases of Models trained in DP-based Fair Learning 6.3 DP-based Fair Classification in Heterogeneous Federated Learning 6.3 DP-based Fair Classification in Heterogeneous Federated Learning 7 Conclusion and References 7 Conclusion and References Appendix A Proofs Appendix A Proofs Appendix B Additional Results for Image Dataset Appendix B Additional Results for Image Dataset 4 Inductive Biases of DP-based Fair Supervised Learning Proof. We defer the proof to the Appendix Proof We remark the difference between the bias levels shown for the DDP case in Theorem 1 and the other dependence metrics in Theorem 2. The bias level for a DDP-based fair leaner could be considerably stronger than that of mutual information, ERMI, and maximal correlation-based fair learners, as the maximum the total variations in Theorem 1 is replaced by their expected value over PS in Theorem 2. S 4.1 Extending the Theoretical Results to Randomized Prediction Rule Proof. We defer the proof to the Appendix. Proof 5 A Distributionally Robust Optimization Approach to DP-based Fair Learning This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv Authors:
(1) Haoyu LEI, Department of Computer Science and Engineering, The Chinese University of Hong Kong (hylei22@cse.cuhk.edu.hk);
(2) Amin Gohari, Department of Information Engineering, The Chinese University of Hong Kong (agohari@ie.cuhk.edu.hk);
(3) Farzan Farnia, Department of Computer Science and Engineering, The Chinese University of Hong Kong (farnia@cse.cuhk.edu.hk). Authors: Authors: (1) Haoyu LEI, Department of Computer Science and Engineering, The Chinese University of Hong Kong (hylei22@cse.cuhk.edu.hk); (2) Amin Gohari, Department of Information Engineering, The Chinese University of Hong Kong (agohari@ie.cuhk.edu.hk); (3) Farzan Farnia, Department of Computer Science and Engineering, The Chinese University of Hong Kong (farnia@cse.cuhk.edu.hk).

The Limits of Demographic Parity in AI Models 

Table of Links Abstract and 1. Introduction Abstract and 1. Introduction 2 Think-and-Execute 2 Think-and-Execute 3 Experimental Setup 3 Experimental Setup 4 Results 4 Results 5 Analysis 5 Analysis 6 Related Work 6 Related Work 7 Limitations and Discussion 7 Limitations and Discussion 8 Conclusion and References 8 Conclusion and References A Experimental Details A Experimental Details B Details of Think-and-Execute B Details of Think-and-Execute C Prompts Used in Our Experiments C Prompts Used in Our Experiments D Human-written Pseudocode Prompts D Human-written Pseudocode Prompts E Generated Analyses E Generated Analyses F Generated Pseudocode Prompts F Generated Pseudocode Prompts G Qualitative Analysis G Qualitative Analysis 7 Limitations and Discussion A possible limitation of our approach is that we focus on algorithmic reasoning, as we believe it is the best setting to assess LLMs’ capabilities in understanding a complex logic and carrying out a sequence of reasoning step, following the logic. However, we believe that THINK-AND-EXECUTE can be applied to other domains of reasoning that require following a long sequence of reasoning steps, such as multi-hop reasoning (Ji et al., 2020) and symbolic reasoning (Madaan & Yazdanbakhsh, 2022). This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. available on arxiv Authors:
(1) Hyungjoo Chae, Yonsei University;
(2) Yeonghyeon Kim, Yonsei University;
(3) Seungone Kim, KAIST AI;
(4) Kai Tzu-iunn Ong, Yonsei University;
(5) Beong-woo Kwak, Yonsei University;
(6) Moohyeon Kim, Yonsei University;
(7) Seonghwan Kim, Yonsei University;
(8) Taeyoon Kwon, Yonsei University;
(9) Jiwan Chung, Yonsei University;
(10) Youngjae Yu, Yonsei University;
(11) Jinyoung Yeo, Yonsei University. Authors: Authors: (1) Hyungjoo Chae, Yonsei University; (2) Yeonghyeon Kim, Yonsei University; (3) Seungone Kim, KAIST AI; (4) Kai Tzu-iunn Ong, Yonsei University; (5) Beong-woo Kwak, Yonsei University; (6) Moohyeon Kim, Yonsei University; (7) Seonghwan Kim, Yonsei University; (8) Taeyoon Kwon, Yonsei University; (9) Jiwan Chung, Yonsei University; (10) Youngjae Yu, Yonsei University; (11) Jinyoung Yeo, Yonsei University.

Think-and-Execute: The Limitations That We Face 

Author:
(1) Thierry Boy de la Tour, Univ. Grenoble Alpes, CNRS, Grenoble INP, LIG 38000 Grenoble, France. Author: Author: (1) Thierry Boy de la Tour, Univ. Grenoble Alpes, CNRS, Grenoble INP, LIG 38000 Grenoble, France. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Basic Definitions and Notations 2.1 Sets 2.1 Sets 2.2 Sequences 2.2 Sequences 2.3 Signatures and Algebras and 2.4 Categories 2.3 Signatures and Algebras and 2.4 Categories 3 Monographs and their Morphisms 3 Monographs and their Morphisms 4 Limits and Colimits 4 Limits and Colimits 5 Drawing Monographs 5 Drawing Monographs 6 Graph Structures and Typed Monographs 6 Graph Structures and Typed Monographs 7 Submonographs and Partial Morphisms 7 Submonographs and Partial Morphisms 8 Algebraic Transformations of Monographs 8 Algebraic Transformations of Monographs 9 Attributed Typed Monographs 9 Attributed Typed Monographs 10 Conclusion and References 10 Conclusion and References 4 Limits and Colimits The colimits of monographs follow the standard constructions of colimits in Sets and Graphs. Sets Graphs Theorem 4.4. The categories of Definition 3.5 are finitely co-complete. Theorem 4.4. The categories of Definition 3.5 are finitely co-complete. We next investigate the limits in categories of monographs. Products of monographs are more difficult to build than products of graphs. This is due to the fact that edges of identical length may be adjacent to edges of different lengths. We therefore see that E (A x B) is only a subset of EA x EB. Lemma 4.6. Every pair of parallel morphisms f, g : A → B has an equalizer (E, e) eq such that E is finite whenever A is finite. Lemma 4.6. Every pair of parallel morphisms f, g : A → B has an equalizer (E, e) eq such that E is finite whenever A is finite. Corollary 4.7. The monomorphisms in Monogr are the injective morphisms. Corollary 4.7. The monomorphisms in Monogr are the injective morphisms. A well-known consequence of Lemmas 4.5 and 4.6 is that all non-empty finite diagrams in Monogr have limits. Since a limit of O-monographs (resp. standard monographs) is an O-monograph (resp. standard), this holds for all categories of Definition 3.5. In particular they all have pullbacks. We shall now investigate the limits of the empty diagram in these categories, i.e., their possible terminal objects. Definition 4.8. For any set of ordinals O, let Definition 4.8. For any set of ordinals O, let Lemma 4.11. Monogr, SMonogr and FMonogr have no terminal object. Lemma 4.11. Monogr, SMonogr and FMonogr have no terminal object. Since terminal objects are limits of empty diagrams obviously these categories are not finitely complete. Theorem 4.12. O-SMonogr is finitely complete for every set of ordinals O. The categories Monogr, SMonogr and FMonogr are not finitely complete. Theorem 4.12. O SMonogr is finitely complete for every set of ordinals O. The categories Monogr SMonogr and FMonogr are not finitely complete. Proof. By Lemmas 4.5, 4.6, 4.9 and 4.11. Proof The category Graphs is also known to be adhesive, a property of pushouts and pullbacks that has important consequences on algebraic transformations (see [8]) and that we shall therefore investigate. Graphs Definition 4.13 (van Kampen squares, adhesive categories). A pushout square (A, B, C, D) is a van Kampen square if for any commutative cube Definition 4.13 A pushout square if for any commutative cube A category has pushouts along monomorphisms if all sources (A, f, g) have pushouts whenever f or g is a monomorphism. A category has pushouts along monomorphisms if all sources (A, f, g) have pushouts whenever f or g is a monomorphism. A category is adhesive if it has pullbacks, pushouts along monomorphisms and all such pushouts are van Kampen squares. A category is adhesive if it has pullbacks, pushouts along monomorphisms and all such pushouts are van Kampen squares. As in the proof that Graphs is adhesive, we will use the fact that the category Sets is adhesive. Graphs Sets Lemma 4.14. E reflects isomorphisms. Lemma 4.14. reflects isomorphisms. A side consequence is that Monogr is balanced, i.e., if f is both a monomorphism and an epimorphism, then by Corollaries 4.3 and 4.7 f is bijective, hence is an isomorphism. More important is that we can use [7, Theorem 24.7], i.e., that a faithful and isomorphism reflecting functor from a category that has some limits or colimits and preserves them, also reflects them. Monogr Lemma 4.15. E preserves and reflects finite colimits. Lemma 4.15. preserves and reflects finite colimits. Proof. It is easy to see from the proofs of Lemmas 4.1 and 4.2 that E preserves both coproducts and coequalizers, so that E preserves all finite co-limits and hence also reflects them. Proof This is particularly true for pushouts. The situation for pullbacks is more complicated since E does not preserve products. Lemma 4.16. E preserves and reflects pullbacks. preserves and reflects pullbacks. Theorem 4.17. The categories of Definition 3.5 are adhesive. Theorem 4.17. The categories of Definition 3.5 are adhesive. Proof. The existence of pullbacks and pushouts is already established. In any of these categories a commutative cube built on a pushout along a monomorphism as bottom face and with pullbacks as back faces, has an underlying cube in Sets that has the same properties by Corollary 4.7, Lemmas 4.15 and 4.16. Since Sets is an adhesive category (see [8]) the underlying bottom face is a van Kampen square, hence such is the bottom face of the initial cube by Lemmas 4.15 and 4.16. Proof Sets Sets This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

A Structural Analysis of Limits and Colimits in Monographs 

Being a software engineer who works extensively with financial data, I recently hit a wall with traditional stock market APIs. After getting frustrated with rate limits and expensive subscriptions, I decided to build my own solution using web scraping. Here's how I did it, and what I learned along the way. Introduction: Why I Needed a Different Approach My breaking point came during a personal project where I was trying to analyze market trends. Yahoo Finance's API kept hitting rate limits, and Bloomberg Terminal's pricing made me laugh out loud - there was no way I could justify that cost for a side project. I needed something that would let me: Fetch data without arbitrary limits
Get real-time prices and trading volumes
Access historical data without paying premium fees
Scale up my analysis as needed The Web Scraping Solution After some research and experimentation, I settled on scraping data from two main sources: CNN Money for trending stocks and Yahoo Finance for detailed metrics. Here's how I built it: Setting Up the Basic Infrastructure First, I installed the essential tools: pip install requests bs4 Then I created a basic scraper that could handle network issues gracefully: import requests
from bs4 import BeautifulSoup
import time
import logging

def make_request(url, max_retries=3):
    headers = {
        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
    }
    
    for attempt in range(max_retries):
        try:
            return requests.get(url, headers=headers, timeout=10)
        except Exception as e:
            if attempt == max_retries - 1:
                raise
            time.sleep(attempt + 1) Grabbing Trending Stocks I started with CNN Money's hot stocks list, which gives me three categories of stocks to track: def get_trending_stocks():
    url = 'https://money.cnn.com/data/hotstocks/index.html'
    response = make_request(url)
    soup = BeautifulSoup(response.text, "html.parser")

    tables = soup.findAll("table", {"class": "wsod_dataTable wsod_dataTableBigAlt"})
    categories = ["Most Actives", "Gainers", "Losers"]
    
    stocks = []
    for i, table in enumerate(tables):
        for row in table.findAll("tr")[1:]:  # Skip headers
            cells = row.findAll("td")
            if cells:
                stocks.append({
                    'category': categories[i],
                    'symbol': cells[0].find(text=True),
                    'company': cells[0].span.text.strip()
                })
    
    return stocks Getting the Financial Details For each trending stock, I fetch additional data from Yahoo Finance: def get_stock_details(symbol):
    url = f"https://finance.yahoo.com/quote/{symbol}"
    response = make_request(url)
    soup = BeautifulSoup(response.text, "html.parser")

    data = {}
    
    # Find the main quote table
    table = soup.find("table", {"class": "W(100%)"})
    if table:
        for row in table.findAll("tr"):
            cells = row.findAll("td")
            if len(cells) > 1:
                key = cells[0].text.strip()
                value = cells[1].text.strip()
                data[key] = value
    
    return data The Gotchas I Encountered Building this wasn't all smooth sailing. Here are some real issues I hit and how I solved them: Rate Limiting: Yahoo Finance started blocking me after too many rapid requests. I added random delays between requests: time.sleep(random.uniform(1, 3))  # Random delay between 1-3 seconds Data Inconsistencies: Sometimes the scraped data would be malformed. I added validation: def validate_price(price_str):
    try:
        return float(price_str.replace('$', '').replace(',', ''))
    except:
        return None Website Changes: The sites occasionally update their HTML structure. I made my selectors more robust: # Instead of exact class matches, use partial matches
table = soup.find("table", class_=lambda x: x and 'dataTable' in x) Storing and Using the Data I keep things simple with CSV storage - it's easy to work with and perfect for my needs: import csv
from datetime import datetime

def save_stock_data(stocks):
    timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
    
    with open('stock_data.csv', 'a', newline='') as file:
        writer = csv.writer(file)
        for stock in stocks:
            writer.writerow([timestamp, stock['symbol'], 
                           stock['price'], stock['volume']]) What I Learned After running this scraper for several weeks, here are my key takeaways: Web scraping isn't just a hack - it's a viable alternative to expensive APIs when done rightly.
Building in error handling and logging from the start saves huge headaches later.
Stock data is messy - so always validate what you scrape.
Starting simple and iterating works better than trying to build everything at once! What's Next? I'm currently working on adding: News sentiment analysis
Basic pattern recognition
A simple dashboard for visualization Also, would you like to integrate this scraper with machine learning models to predict stock trends? Let me know in the comments! Being a software engineer who works extensively with financial data, I recently hit a wall with traditional stock market APIs. After getting frustrated with rate limits and expensive subscriptions, I decided to build my own solution using web scraping. Here's how I did it, and what I learned along the way. Introduction: Why I Needed a Different Approach My breaking point came during a personal project where I was trying to analyze market trends. Yahoo Finance's API kept hitting rate limits, and Bloomberg Terminal's pricing made me laugh out loud - there was no way I could justify that cost for a side project. I needed something that would let me: Fetch data without arbitrary limits Get real-time prices and trading volumes Access historical data without paying premium fees Scale up my analysis as needed Fetch data without arbitrary limits Get real-time prices and trading volumes Access historical data without paying premium fees Scale up my analysis as needed The Web Scraping Solution After some research and experimentation, I settled on scraping data from two main sources: CNN Money for trending stocks and Yahoo Finance for detailed metrics. Here's how I built it: Setting Up the Basic Infrastructure First, I installed the essential tools: pip install requests bs4 pip install requests bs4 Then I created a basic scraper that could handle network issues gracefully: import requests
from bs4 import BeautifulSoup
import time
import logging

def make_request(url, max_retries=3):
    headers = {
        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
    }
    
    for attempt in range(max_retries):
        try:
            return requests.get(url, headers=headers, timeout=10)
        except Exception as e:
            if attempt == max_retries - 1:
                raise
            time.sleep(attempt + 1) import requests
from bs4 import BeautifulSoup
import time
import logging

def make_request(url, max_retries=3):
    headers = {
        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
    }
    
    for attempt in range(max_retries):
        try:
            return requests.get(url, headers=headers, timeout=10)
        except Exception as e:
            if attempt == max_retries - 1:
                raise
            time.sleep(attempt + 1) Grabbing Trending Stocks I started with CNN Money's hot stocks list, which gives me three categories of stocks to track: def get_trending_stocks():
    url = 'https://money.cnn.com/data/hotstocks/index.html'
    response = make_request(url)
    soup = BeautifulSoup(response.text, "html.parser")

    tables = soup.findAll("table", {"class": "wsod_dataTable wsod_dataTableBigAlt"})
    categories = ["Most Actives", "Gainers", "Losers"]
    
    stocks = []
    for i, table in enumerate(tables):
        for row in table.findAll("tr")[1:]:  # Skip headers
            cells = row.findAll("td")
            if cells:
                stocks.append({
                    'category': categories[i],
                    'symbol': cells[0].find(text=True),
                    'company': cells[0].span.text.strip()
                })
    
    return stocks def get_trending_stocks():
    url = 'https://money.cnn.com/data/hotstocks/index.html'
    response = make_request(url)
    soup = BeautifulSoup(response.text, "html.parser")

    tables = soup.findAll("table", {"class": "wsod_dataTable wsod_dataTableBigAlt"})
    categories = ["Most Actives", "Gainers", "Losers"]
    
    stocks = []
    for i, table in enumerate(tables):
        for row in table.findAll("tr")[1:]:  # Skip headers
            cells = row.findAll("td")
            if cells:
                stocks.append({
                    'category': categories[i],
                    'symbol': cells[0].find(text=True),
                    'company': cells[0].span.text.strip()
                })
    
    return stocks Getting the Financial Details For each trending stock, I fetch additional data from Yahoo Finance: def get_stock_details(symbol):
    url = f"https://finance.yahoo.com/quote/{symbol}"
    response = make_request(url)
    soup = BeautifulSoup(response.text, "html.parser")

    data = {}
    
    # Find the main quote table
    table = soup.find("table", {"class": "W(100%)"})
    if table:
        for row in table.findAll("tr"):
            cells = row.findAll("td")
            if len(cells) > 1:
                key = cells[0].text.strip()
                value = cells[1].text.strip()
                data[key] = value
    
    return data def get_stock_details(symbol):
    url = f"https://finance.yahoo.com/quote/{symbol}"
    response = make_request(url)
    soup = BeautifulSoup(response.text, "html.parser")

    data = {}
    
    # Find the main quote table
    table = soup.find("table", {"class": "W(100%)"})
    if table:
        for row in table.findAll("tr"):
            cells = row.findAll("td")
            if len(cells) > 1:
                key = cells[0].text.strip()
                value = cells[1].text.strip()
                data[key] = value
    
    return data The Gotchas I Encountered Building this wasn't all smooth sailing. Here are some real issues I hit and how I solved them: Rate Limiting: Yahoo Finance started blocking me after too many rapid requests. I added random delays between requests: Rate Limiting : Yahoo Finance started blocking me after too many rapid requests. I added random delays between requests: Rate Limiting time.sleep(random.uniform(1, 3))  # Random delay between 1-3 seconds time.sleep(random.uniform(1, 3))  # Random delay between 1-3 seconds Data Inconsistencies: Sometimes the scraped data would be malformed. I added validation: Data Inconsistencies : Sometimes the scraped data would be malformed. I added validation: Data Inconsistencies def validate_price(price_str):
    try:
        return float(price_str.replace('$', '').replace(',', ''))
    except:
        return None def validate_price(price_str):
    try:
        return float(price_str.replace('$', '').replace(',', ''))
    except:
        return None Website Changes: The sites occasionally update their HTML structure. I made my selectors more robust: Website Changes : The sites occasionally update their HTML structure. I made my selectors more robust: Website Changes # Instead of exact class matches, use partial matches
table = soup.find("table", class_=lambda x: x and 'dataTable' in x) # Instead of exact class matches, use partial matches
table = soup.find("table", class_=lambda x: x and 'dataTable' in x) Storing and Using the Data I keep things simple with CSV storage - it's easy to work with and perfect for my needs: import csv
from datetime import datetime

def save_stock_data(stocks):
    timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
    
    with open('stock_data.csv', 'a', newline='') as file:
        writer = csv.writer(file)
        for stock in stocks:
            writer.writerow([timestamp, stock['symbol'], 
                           stock['price'], stock['volume']]) import csv
from datetime import datetime

def save_stock_data(stocks):
    timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
    
    with open('stock_data.csv', 'a', newline='') as file:
        writer = csv.writer(file)
        for stock in stocks:
            writer.writerow([timestamp, stock['symbol'], 
                           stock['price'], stock['volume']]) What I Learned After running this scraper for several weeks, here are my key takeaways: After running this scraper for several weeks, here are my key takeaways: Web scraping isn't just a hack - it's a viable alternative to expensive APIs when done rightly. Building in error handling and logging from the start saves huge headaches later. Stock data is messy - so always validate what you scrape. Starting simple and iterating works better than trying to build everything at once! Web scraping isn't just a hack - it's a viable alternative to expensive APIs when done rightly. Building in error handling and logging from the start saves huge headaches later. Stock data is messy - so always validate what you scrape. Starting simple and iterating works better than trying to build everything at once! What's Next? I'm currently working on adding: I'm currently working on adding: News sentiment analysis Basic pattern recognition A simple dashboard for visualization News sentiment analysis Basic pattern recognition A simple dashboard for visualization Also, would you like to integrate this scraper with machine learning models to predict stock trends? Let me know in the comments! machine learning models

How to Build a No-Limits Stock Market Scraper with Python

Table of Links Abstract and I. Introduction II. Background III. Related Works IV. Comparative Analysis V. Findings and Evaluation VI. Conclusion and References V. FINDINGS AND EVALUATION Through this research we’ve observed the inability to successfully implement layer-one sharding on account based transactional models. As demonstrated in section III. B account based systems implement layer-two scaling to emulate the parallelization of UTXO sharding. We also observe the inability for off-chain automation dependent solutions to exist on UTXO based systems, due to smart contract incompatibility. The impact of this lack of solution interoperability between layer one and two presents a glass ceiling of scale out potential. Layer two projects such as Plasma still rely on a heavily constrained main chain for coordination and dispute resolution, zero-knowledge and optimistic rollups all have similar dependencies on parent coordination. If sharding were implemented harmoniously underneath these architectures, bottlenecks would be avoided in main-chain coordination. This allows both scalability layers to scale-out by a factor of the other. Imagine a scenario where a mainchain can handle 10,000 transactions-per-second through off-chain scaling and smart contract automation. The overall network architecture can continually scale on layer-two through nested parent/child relationships, however coordination and resolution cause a mainchain bottleneck. Resolving this with layer-one solutions, such as OmniLedgers linear scale-out sharding would increase potential speeds by a factor of n, where n is equal to mainchain parrelization. To address this we propose the use of an Extended-UTXO model that allows for smart contract compatibility. This specification allows for off-chain automation in rollups, while maintaining the ability to shard UTXO states and transactions at layer-one. A. Design Architecture The eUTXO follows the same graph based approach that the UTXO model does, complementing the concurrent and distributed attributes of blockchains. The graph based model eliminates the need for a globally shared mutable state, which is known to cause highly complex problems when facing computations involving that shared state [6]. The first account model blockchain, Ethereum, did solve how to run expressive smart contracts on a global state machine capable of execution on-chain. The eUTXO is an extension onto the UTXO ledger which notably increases the expressiveness of the model, while still utilizing the concurrent data flows of unspent transactions. In creating a more extensible model, transactions can occur on more expressive state machines without needing to use an account model. Each individual transaction that occurs on a concurrent chain represents steps in state machine and to utilize a more expressive model, the state needs to be maintained and the same contract code is used along the whole sequence of transactions. The eUTXO model maintains contract state by extending unspent transaction outputs from being a pair of validator and digital asset to a now being a treble containing a validator, the asset and a datum, where the datum contains arbitrary contract specific data [6]. The eUTXO model implements local state machines which are capable of running smart contracts on them to produces new unspent transactions as well as maintain the new state to be maintained. This enables developers to build distributed applications on massively scalable platforms. OmniLedger inspired sharding protocols will be adapted to the eUTXO model using verifiable random function (VRF) to accomplish network sharding at layer-one [20]. We intend to use a temporary shard coordinator delegated through VRF election, to generate an unbiased random string. Each epoch the unbiased random string is used in shard assignment. eUTXO transactions will be assigned through the use of a randomly chosen directory service shard. This phase is based on Zilliqa’s approach to transaction sharding [22]. We intend to use the unbiased random string generated by the temporary epoch coordinator to delegate this directory service shard. The directory shard will assign transactions to shards based on the sender’s address and smart contract state being interacted with. Smart contract states will be maintained in the eUTXO transactions respective shard. Contracts will be fully executed globally at the end of each epoch when the directory service coordinator synchronises the entire network state. As the eUTXO model supports the use of smart contracts, further layer 2 solutions such as Zero-Knowledge and Optimistic Rollups could be implemented, and even create nestedchains that communicate through smart contract two-way pegs which could increase throughput exponentially. Authors:
(1) Cayo Fletcher-Smith, Bournemouth University;
(2) Frazer Chard. This paper is available on arxiv under CC0 1.0 DEED license. Table of Links Abstract and I. Introduction Abstract and I. Introduction II. Background II. Background III. Related Works III. Related Works IV. Comparative Analysis IV. Comparative Analysis V. Findings and Evaluation V. Findings and Evaluation VI. Conclusion and References VI. Conclusion and References V. FINDINGS AND EVALUATION Through this research we’ve observed the inability to successfully implement layer-one sharding on account based transactional models. As demonstrated in section III. B account based systems implement layer-two scaling to emulate the parallelization of UTXO sharding. We also observe the inability for off-chain automation dependent solutions to exist on UTXO based systems, due to smart contract incompatibility. The impact of this lack of solution interoperability between layer one and two presents a glass ceiling of scale out potential. Layer two projects such as Plasma still rely on a heavily constrained main chain for coordination and dispute resolution, zero-knowledge and optimistic rollups all have similar dependencies on parent coordination. If sharding were implemented harmoniously underneath these architectures, bottlenecks would be avoided in main-chain coordination. This allows both scalability layers to scale-out by a factor of the other. Imagine a scenario where a mainchain can handle 10,000 transactions-per-second through off-chain scaling and smart contract automation. The overall network architecture can continually scale on layer-two through nested parent/child relationships, however coordination and resolution cause a mainchain bottleneck. Resolving this with layer-one solutions, such as OmniLedgers linear scale-out sharding would increase potential speeds by a factor of n, where n is equal to mainchain parrelization. To address this we propose the use of an Extended-UTXO model that allows for smart contract compatibility. This specification allows for off-chain automation in rollups, while maintaining the ability to shard UTXO states and transactions at layer-one. A. Design Architecture A. Design Architecture The eUTXO follows the same graph based approach that the UTXO model does, complementing the concurrent and distributed attributes of blockchains. The graph based model eliminates the need for a globally shared mutable state, which is known to cause highly complex problems when facing computations involving that shared state [6]. The first account model blockchain, Ethereum, did solve how to run expressive smart contracts on a global state machine capable of execution on-chain. The eUTXO is an extension onto the UTXO ledger which notably increases the expressiveness of the model, while still utilizing the concurrent data flows of unspent transactions. In creating a more extensible model, transactions can occur on more expressive state machines without needing to use an account model. Each individual transaction that occurs on a concurrent chain represents steps in state machine and to utilize a more expressive model, the state needs to be maintained and the same contract code is used along the whole sequence of transactions. The eUTXO model maintains contract state by extending unspent transaction outputs from being a pair of validator and digital asset to a now being a treble containing a validator, the asset and a datum, where the datum contains arbitrary contract specific data [6]. The eUTXO model implements local state machines which are capable of running smart contracts on them to produces new unspent transactions as well as maintain the new state to be maintained. This enables developers to build distributed applications on massively scalable platforms. OmniLedger inspired sharding protocols will be adapted to the eUTXO model using verifiable random function (VRF) to accomplish network sharding at layer-one [20]. We intend to use a temporary shard coordinator delegated through VRF election, to generate an unbiased random string. Each epoch the unbiased random string is used in shard assignment. eUTXO transactions will be assigned through the use of a randomly chosen directory service shard. This phase is based on Zilliqa’s approach to transaction sharding [22]. We intend to use the unbiased random string generated by the temporary epoch coordinator to delegate this directory service shard. The directory shard will assign transactions to shards based on the sender’s address and smart contract state being interacted with. Smart contract states will be maintained in the eUTXO transactions respective shard. Contracts will be fully executed globally at the end of each epoch when the directory service coordinator synchronises the entire network state. As the eUTXO model supports the use of smart contracts, further layer 2 solutions such as Zero-Knowledge and Optimistic Rollups could be implemented, and even create nestedchains that communicate through smart contract two-way pegs which could increase throughput exponentially. Authors: (1) Cayo Fletcher-Smith, Bournemouth University; (2) Frazer Chard. Authors: Authors: (1) Cayo Fletcher-Smith, Bournemouth University; (2) Frazer Chard. This paper is available on arxiv under CC0 1.0 DEED license. This paper is available on arxiv under CC0 1.0 DEED license. available on arxiv

New Research Reveals How Blockchain Can Scale Beyond Current Limits with eUTXO

Authors:
(1) Avrim Blum, Toyota Technological Institute at Chicago, IL, USA;
(2) Melissa Dutz, Toyota Technological Institute at Chicago, IL, USA. Table of Links Abstract and 1 Introduction 2 Setting and 2.1 Models of behaviorally-biased opponents 3 Preliminaries and Intuition 4.1 Myopic Best Responder and 4.2 Gambler’s Fallacy Opponent 4.3 Win-Stay, Lose-Shift Opponent 4.4 Follow-the-Leader Opponent and 4.5 Highest Average Payoff Opponent 5 Generalizing 5.1 Other Behaviorally-Biased Strategies 5.2 Exploiting an Unknown Strategy from a Known Set of Strategies 6 Future Work and References A Appendix A.1 Win-Stay Lose-Shift Variant: Tie-Stay A.2 Follow-the-Leader Variant: Limited History A.3 Ellipsoid Mistake Bounds A.4 Highest Average Payoff Opponent A.2 Follow-the-Leader Variant: Limited History Recall that the limited-history variant of the Follow-the-Leader opponent plays the action that would have achieved the highest net payoff against the last r rounds of our play. Note that we assume r > 0. If r = 0, the opponent would simply play the same action every round (the first action in their action ordering) regardless of our play. In that case, their play would not reveal any information about best responses, so the best we could do is tie in most rounds by playing the same action as the opponent. We again use the ellipsoid algorithm for prediction, essentially the same way as described for the unlimited-history variant of this opponent in 4.4. In this case, the inequality we observe when a mistake is made is only relative to the past r rounds. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Avrim Blum, Toyota Technological Institute at Chicago, IL, USA; (2) Melissa Dutz, Toyota Technological Institute at Chicago, IL, USA. Authors: Authors: (1) Avrim Blum, Toyota Technological Institute at Chicago, IL, USA; (2) Melissa Dutz, Toyota Technological Institute at Chicago, IL, USA. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Setting and 2.1 Models of behaviorally-biased opponents 2 Setting and 2.1 Models of behaviorally-biased opponents 3 Preliminaries and Intuition 3 Preliminaries and Intuition 4.1 Myopic Best Responder and 4.2 Gambler’s Fallacy Opponent 4.1 Myopic Best Responder and 4.2 Gambler’s Fallacy Opponent 4.3 Win-Stay, Lose-Shift Opponent 4.3 Win-Stay, Lose-Shift Opponent 4.4 Follow-the-Leader Opponent and 4.5 Highest Average Payoff Opponent 4.4 Follow-the-Leader Opponent and 4.5 Highest Average Payoff Opponent 5 Generalizing 5.1 Other Behaviorally-Biased Strategies 5.1 Other Behaviorally-Biased Strategies 5.2 Exploiting an Unknown Strategy from a Known Set of Strategies 5.2 Exploiting an Unknown Strategy from a Known Set of Strategies 6 Future Work and References 6 Future Work and References A Appendix A.1 Win-Stay Lose-Shift Variant: Tie-Stay A.1 Win-Stay Lose-Shift Variant: Tie-Stay A.2 Follow-the-Leader Variant: Limited History A.2 Follow-the-Leader Variant: Limited History A.3 Ellipsoid Mistake Bounds A.3 Ellipsoid Mistake Bounds A.4 Highest Average Payoff Opponent A.4 Highest Average Payoff Opponent A.2 Follow-the-Leader Variant: Limited History Recall that the limited-history variant of the Follow-the-Leader opponent plays the action that would have achieved the highest net payoff against the last r rounds of our play. Note that we assume r > 0. If r = 0, the opponent would simply play the same action every round (the first action in their action ordering) regardless of our play. In that case, their play would not reveal any information about best responses, so the best we could do is tie in most rounds by playing the same action as the opponent. We again use the ellipsoid algorithm for prediction, essentially the same way as described for the unlimited-history variant of this opponent in 4.4. In this case, the inequality we observe when a mistake is made is only relative to the past r rounds. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

Ways to Counter Limited-History Opponents with Algorithmic Tools

To Pump The Gas Or Not: Analyzing The Ethereum Gas Limit Debate

Qazı Pompalamaq və ya etməmək: Ethereum Qaz Limiti Müzakirəsinin Təhlili

Pumpati plin ili ne: Analiza debate o ograničenju Ethereum plina

At pumpe gassen eller ej: Analyse af Ethereum-gasgrænsedebatten

Bombear gasolina o no: análisis del debate sobre el límite de gas de Ethereum

ガスを注入するか否か: イーサリアムのガス制限論争を分析

Gas pompen of niet: analyse van het Ethereum-gaslimietdebat

Kuvoma gaze cyangwa Oya: Gusesengura Impaka za Gaz ya Ethereum

ගෑස් පොම්ප කිරීමට හෝ නොකිරීමට: Ethereum ගෑස් සීමාව විවාදය විශ්ලේෂණය කිරීම

Kupomba Gasi Kana Kuti Kwete: Kuongorora Iyo Ethereum Gasi Limit Gakava

Për të pompuar gazin apo jo: duke analizuar debatin e kufirit të gazit Ethereum

Ukumpompa igesi okanye hayi: Ukuhlalutya iNgxoxo yoMda weGesi ye-Ethereum

Ukupompa Igesi Noma Cha: Ukuhlaziya Inkulumo-mpikiswano Yomkhawulo Wegesi ye-Ethereum

Chrome extensions are a great way for developers to provide users with useful tools without having to build an entire web app. However, Chrome extensions have a specific structure, and data persistence can be tricky. While there are great docs for Chrome’s Storage API, that doesn’t cover everything. Currently, no single resource clearly explains all the options for persisting state in Chrome extensions for different durations. This guide explains the options and limitations for state persistence in Chrome extensions, helping developers choose the best fit for their projects. Popup lifetime storage - the default for popups For Chrome extensions that use a popup, anytime the user opens a popup, it is a new instance of that popup. So if the user closes the popup (which happens automatically if they click into the browser), and then reopens it, that new popup will not remember or be aware of anything done previously. Side Panel lifetime storage - default for side panels Like popups, when the user closes and reopens a side panel, they are creating a new instance of the side panel.  Data will only persist as long as the side panel is open, and any data created when they had the side panel open earlier will be lost. However, unlike popups, side panels will not automatically close whenever the user clicks outside of it, so side panels are able to have much longer lifetimes than popups. A side panel can stay open as the user navigates to different web pages in a tab or even different tabs within Chrome. As a result, data in a side panel is capable of persisting much longer than in a popup, though it still will be lost if the user closes and reopens it. Page Lifetime Storage - Message Passing If an extension interacts with the webpage, the extension must track changes or actions already taken. For instance, if an extension changes a web page’s colors to display in dark mode, it should remember this, so it does not prompt the user to repeat something they have already done. Additionally, if the user refreshes the page, any previously executed code is cleared. So, we don’t want to persist data beyond the current page session, as it will be outdated. To handle this, we use message passing. The content script has the code we run on the webpage, which automatically persists for the page lifetime. However, as mentioned above, popups and side panels - the UIs for the user - do not. Message passing allows us to make sure the UIs are up to date. First, we define the state in the content script (e.g., "initial," "loading," "success," or "error"). We pass a message every time it changes so the UI is up to date. We also request the current state every time the popup or side panel is opened. So if the user closes and reopens a popup while the extension is loading, the new popup will request the state, receive ‘loading’, and render the ‘loading’ UI. Here is an example: popup.js chrome.tabs.query({active: true, currentWindow: true}, function(tabs) {
   chrome.tabs.sendMessage(tabs[0].id, {action: "getExtensionState"}, function(response) {
       // if the content script hasn't been injected, then the code in that script hasn't been run, and we'll get an error or no response
       if (chrome.runtime.lastError || !response) {
           return;
       } else if (response) {
           // if the code in content script has been injected, we'll get a response which tells us what state the code is at (initial, loading, success, error, etc)
           renderUI(response.extensionState)
       }
   });
}); content-script.js let extensionState = 'initial';

chrome.runtime.onMessage.addListener(
   function(request, sender, sendResponse) {
     if (request.action === "getExtensionState")
       sendResponse({extensionState});
   }); In some cases, the content script will be injected every time we open the extension. In others, we run the content script at a specific time - ie, the user clicks a button in the popup, which then injects the code (ie,  “Click me to display this page in dark mode”). If the content script has never been run, we will get an error when we send the “getExtensionState” message from the popup. The first if statement in the example popup.js code is there to catch that error and allow us to do whatever we want to do. Deprecated: extension lifetime - background worker This option is deprecated and no longer usable. However, a lot of Chrome resources are older, and it’s not always obvious which 8 year old Stack Overflow answers are still relevant and which ones aren’t, so I’m briefly touching on it so you know not to try to implement this. In past versions of Chrome extensions, there was the option of ‘background workers’ (V2 background worker docs). It was possible to set these to ‘persist,’ and they would stay active for the extension instance lifetime, and across tabs. The field ‘background’ still exists in Manifest V3, but it is now considered a service worker and only stays active as long as it is running code - it is put to sleep within 30 seconds of inactivity (service workers docs). There is no out of the box option to set these service workers as persistent. Chrome Storage API Chrome’s Storage API provides multiple options to persist data longer. To use any of these, you’ll have to add ‘storage’ permissions in the Manifest. The docs for the Chrome Storage API are here and cover the final three options. Browser Session storage - chrome.storage.session The shortest Chrome Storage API option is chrome session storage, which holds data in memory for the duration of the Chrome browser session. In this context, a Chrome browser session is the time period any windows are open with the same Chrome profile. So if a user has two Chrome windows open, closes one but the other remains open, the user is still in the same browser session, and the data will persist. Additionally, if the user refreshes the page or closes the tab, the data will persist. The data will be cleared only if they close all browser windows for that Chrome profile, restart Chrome, or reload the extension. Example implementation from the API docs: chrome.storage.session.set({ key: value }).then(() => {
  console.log("Value was set");
});

chrome.storage.session.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); Extension installation lifetime storage - chrome.storage.local Chrome’s local storage allows the data to be saved as long as the extension is installed. It will persist even if the page is refreshed, or the browser is fully closed and reopened. It is only cleared if the extension is removed or chrome.storage.local.clear() is called. Example implementation from the API docs: chrome.storage.local.set({ key: value }).then(() => {
  console.log("Value is set");
});

chrome.storage.local.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); Storage across chrome devices - chrome.storage.sync
Chrome’s sync storage will sync the saved data to any Chrome browser the user is logged into. If a user does not have sync enabled, it will fall back to local storage behavior. The data is only cleared if the extension is uninstalled or chrome.storage.local.clear() is called. Example implementation from the API docs: chrome.storage.sync.set({ key: value }).then(() => {
  console.log("Value is set");
});

chrome.storage.sync.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); Did I miss anything? Please let me know in the comments! Chrome extensions are a great way for developers to provide users with useful tools without having to build an entire web app. However, Chrome extensions have a specific structure, and data persistence can be tricky. While there are great docs for Chrome’s Storage API, that doesn’t cover everything. Currently, no single resource clearly explains all the options for persisting state in Chrome extensions for different durations. This guide explains the options and limitations for state persistence in Chrome extensions, helping developers choose the best fit for their projects. Chrome extensions Popup lifetime storage - the default for popups Popup lifetime storage - the default for popups Popup lifetime storage - the default for popups For Chrome extensions that use a popup, anytime the user opens a popup, it is a new instance of that popup. So if the user closes the popup (which happens automatically if they click into the browser), and then reopens it, that new popup will not remember or be aware of anything done previously. Side Panel lifetime storage - default for side panels Side Panel lifetime storage - default for side panels Side Panel lifetime storage - default for side panels Like popups, when the user closes and reopens a side panel, they are creating a new instance of the side panel.  Data will only persist as long as the side panel is open, and any data created when they had the side panel open earlier will be lost. However, unlike popups, side panels will not automatically close whenever the user clicks outside of it, so side panels are able to have much longer lifetimes than popups. A side panel can stay open as the user navigates to different web pages in a tab or even different tabs within Chrome. As a result, data in a side panel is capable of persisting much longer than in a popup, though it still will be lost if the user closes and reopens it. Page Lifetime Storage - Message Passing Page Lifetime Storage - Message Passing Page Lifetime Storage - Message Passing Page Lifetime Storage - Message Passing If an extension interacts with the webpage, the extension must track changes or actions already taken. For instance, if an extension changes a web page’s colors to display in dark mode, it should remember this, so it does not prompt the user to repeat something they have already done. Additionally, if the user refreshes the page, any previously executed code is cleared. So, we don’t want to persist data beyond the current page session, as it will be outdated. To handle this, we use message passing. The content script has the code we run on the webpage, which automatically persists for the page lifetime. However, as mentioned above, popups and side panels - the UIs for the user - do not. Message passing allows us to make sure the UIs are up to date. First, we define the state in the content script (e.g., "initial," "loading," "success," or "error"). We pass a message every time it changes so the UI is up to date. We also request the current state every time the popup or side panel is opened. So if the user closes and reopens a popup while the extension is loading, the new popup will request the state, receive ‘loading’, and render the ‘loading’ UI. Here is an example: popup.js popup.js chrome.tabs.query({active: true, currentWindow: true}, function(tabs) {
   chrome.tabs.sendMessage(tabs[0].id, {action: "getExtensionState"}, function(response) {
       // if the content script hasn't been injected, then the code in that script hasn't been run, and we'll get an error or no response
       if (chrome.runtime.lastError || !response) {
           return;
       } else if (response) {
           // if the code in content script has been injected, we'll get a response which tells us what state the code is at (initial, loading, success, error, etc)
           renderUI(response.extensionState)
       }
   });
}); chrome.tabs.query({active: true, currentWindow: true}, function(tabs) {
   chrome.tabs.sendMessage(tabs[0].id, {action: "getExtensionState"}, function(response) {
       // if the content script hasn't been injected, then the code in that script hasn't been run, and we'll get an error or no response
       if (chrome.runtime.lastError || !response) {
           return;
       } else if (response) {
           // if the code in content script has been injected, we'll get a response which tells us what state the code is at (initial, loading, success, error, etc)
           renderUI(response.extensionState)
       }
   });
}); content-script.js content-script.js let extensionState = 'initial';

chrome.runtime.onMessage.addListener(
   function(request, sender, sendResponse) {
     if (request.action === "getExtensionState")
       sendResponse({extensionState});
   }); let extensionState = 'initial';

chrome.runtime.onMessage.addListener(
   function(request, sender, sendResponse) {
     if (request.action === "getExtensionState")
       sendResponse({extensionState});
   }); In some cases, the content script will be injected every time we open the extension. In others, we run the content script at a specific time - ie, the user clicks a button in the popup, which then injects the code (ie,  “Click me to display this page in dark mode”). If the content script has never been run, we will get an error when we send the “getExtensionState” message from the popup. The first if statement in the example popup.js code is there to catch that error and allow us to do whatever we want to do. Deprecated: extension lifetime - background worker Deprecated: extension lifetime - background worker Deprecated: extension lifetime - background worker This option is deprecated and no longer usable. However, a lot of Chrome resources are older, and it’s not always obvious which 8 year old Stack Overflow answers are still relevant and which ones aren’t, so I’m briefly touching on it so you know not to try to implement this. In past versions of Chrome extensions, there was the option of ‘background workers’ ( V2 background worker docs ). It was possible to set these to ‘persist,’ and they would stay active for the extension instance lifetime, and across tabs. The field ‘background’ still exists in Manifest V3, but it is now considered a service worker and only stays active as long as it is running code - it is put to sleep within 30 seconds of inactivity ( service workers docs ). There is no out of the box option to set these service workers as persistent. V2 background worker docs service workers docs Chrome Storage API Chrome’s Storage API provides multiple options to persist data longer. To use any of these, you’ll have to add ‘storage’ permissions in the Manifest. The docs for the Chrome Storage API are here and cover the final three options. The docs for the Chrome Storage API are here Browser Session storage - chrome.storage.session Browser Session storage - chrome.storage.session Browser Session storage - chrome.storage.session The shortest Chrome Storage API option is chrome session storage, which holds data in memory for the duration of the Chrome browser session. In this context, a Chrome browser session is the time period any windows are open with the same Chrome profile. So if a user has two Chrome windows open, closes one but the other remains open, the user is still in the same browser session, and the data will persist. Additionally, if the user refreshes the page or closes the tab, the data will persist. The data will be cleared only if they close all browser windows for that Chrome profile, restart Chrome, or reload the extension. Example implementation from the API docs: chrome.storage.session.set({ key: value }).then(() => {
  console.log("Value was set");
});

chrome.storage.session.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); chrome.storage.session.set({ key: value }).then(() => {
  console.log("Value was set");
});

chrome.storage.session.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); Extension installation lifetime storage - chrome.storage.local Extension installation lifetime storage - chrome.storage.local Extension installation lifetime storage - chrome.storage.local Chrome’s local storage allows the data to be saved as long as the extension is installed. It will persist even if the page is refreshed, or the browser is fully closed and reopened. It is only cleared if the extension is removed or chrome.storage.local.clear() is called. Example implementation from the API docs: chrome.storage.local.set({ key: value }).then(() => {
  console.log("Value is set");
});

chrome.storage.local.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); chrome.storage.local.set({ key: value }).then(() => {
  console.log("Value is set");
});

chrome.storage.local.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); Storage across chrome devices - chrome.storage.sync
Chrome’s sync storage will sync the saved data to any Chrome browser the user is logged into. If a user does not have sync enabled, it will fall back to local storage behavior. The data is only cleared if the extension is uninstalled or chrome.storage.local.clear() is called. Storage across chrome devices - chrome.storage.sync Chrome’s sync storage will sync the saved data to any Chrome browser the user is logged into. If a user does not have sync enabled, it will fall back to local storage behavior. The data is only cleared if the extension is uninstalled or chrome.storage.local.clear() is called. Storage across chrome devices - chrome.storage.sync Storage across chrome devices - chrome.storage.sync Chrome’s sync storage will sync the saved data to any Chrome browser the user is logged into. If a user does not have sync enabled, it will fall back to local storage behavior. The data is only cleared if the extension is uninstalled or chrome.storage.local.clear() is called. Example implementation from the API docs: Example implementation from the API docs: chrome.storage.sync.set({ key: value }).then(() => {
  console.log("Value is set");
});

chrome.storage.sync.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); chrome.storage.sync.set({ key: value }).then(() => {
  console.log("Value is set");
});

chrome.storage.sync.get(["key"]).then((result) => {
  console.log("Value is " + result.key);
}); Did I miss anything? Please let me know in the comments!

State Storage in Chrome Extensions: Options, Limits, and Best Practices

Challenges for Life in the Venusian Clouds Include Severe Acidity, Extreme Dryness, & Limited Water

Authors:
(1) Clauvin Almeida, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil;
(2) Marcos Kalinowski, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil;
(3) Anderson Uchoa, Federal University of Ceara (UFC), Itapaje, Brazil;
(4) Bruno Feijo, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil. Table of Links Abstract and 1 Introduction 2. Background and Related Work and 2.1. Gamification 2.2. Game Design Elements and 2.3. Gamification Effects 2.4. Related Work on Gamification Negative Effects 3. Systematic Mapping and 3.1. The Research Questions 3.2. Search Strategy and 3.3. Inclusion and Exclusion Criteria 3.4. Applying the Search Strategy 3.5. Data Extraction 4. Systematic Mapping Results 5. Focus Group: Developer Perception on the Negative Effects of Game Design Elements 5.1. Context and Participant Characterization 5.2. Focus Group Design 5.3. The Developers’ Perception on The Negative Effects 5.4. On the Perceived Usefulness, Ease of use and Intent of Adoption of Mapped Negative Effects 5.5. Participant Feedback 6. Limitations 7. Concluding Remarks 7.1. Future Research Directions Acknowledgements and References 6. Limitations One of the limitations of our mapping study concerns potentially missing papers. After analyzing 3458 papers (see the list in our online repository), based on our inclusion and exclusion criteria, we initially included 79 papers. While our combined search strategy allowed identifying significantly more papers than the database search strategy employed by Toda et al. [27] (we identified 34 papers published between 2012 and 2016, against 17), the sets had differences. Therefore, we manually included the papers found by Toda et al., which were missed by our first search and extended searches, ending up with a final set of 87 papers. It is noteworthy that we verified that many of the missed papers would also have been retrieved by subsequent snowballing iterations. Nevertheless, an extension applying subsequent snowballing iterations and investigating different hybrid strategies [15], would require significant additional effort, involving analyzing several thousands of papers which could characterize a different publication instead of an extension. We are confident that our final set of included papers as part of this publication allowed providing an unbiased and meaningful overview of the adverse effects related to GDEs in gamified education software. Another risk of false negatives concerns the filtering process. We screened all papers considering only titles, abstracts, and keywords, which may not contain sufficient information to decide upon inclusion. We avoided applying EC1 and EC3 during the initial screening to lower this risk, only excluding papers that we had high confidence of not investigating GDE effects (EC1) and reporting negative effects (EC3). In case of any doubt, the paper was left for full text-based assessment. Moreover, the application of the inclusion and exclusion criteria made by the first author was reviewed by the second and fourth authors in meetings. In case of doubt, during the initial screening or full-text-based assessment, discussions were held to reach a consensus. Furthermore, we chose not to consider grey literature as part of our inclusion criteria. There is the possibility of relevant grey literature that does not have equivalent in non-grey academic papers. On the other hand, even though we did not explicitly evaluate the strength of evidence, the results herein reported are based on peer-reviewed research and backed by empirical studies. Moreover, while research that reaches negative results is essential because it shows us what does not work [115], there still seems to be a publication bias towards positive outcomes. Research reporting negative results tends to have less scientific interest, fewer citations, and to be less often published [116]. Hence, there may be additional negative results that were not published and which, for that reason, could not be included in our mapping study. Finally, in our summary of the negative effects of GDEs (cf. Figure 4), we included negative effects that may affect different roles (e.g., the students and the maintainers/teachers responsible for configuring the system and keeping it up and running with appropriate educational resources), as we found them reported in the literature. Our goal was to organize evidence regarding the overall adverse effects of GDEs in gamified education software. While we included an overview of the negative effects for different roles (e.g., RQ1.B and RQ1.C), we considered a fine-grained analysis of the effects of each GDEs for each of these roles out of the scope of our intended overview and part of future work. With regard to the focus group session, we wanted to gather insights on the mapped negative effects of game design elements from the point of view of practitioners. In a virtual focus group, generally, the number of participants is reduced, four being considered appropriate (our case) [117]. However, it is suggested to plan more online focus groups with fewer participants than when conducting face-to-face groups [118]. While we carefully planned our focus group, followed best practices, and carefully conducted our qualitative analyses, a single session is surely a limitation. Therefore, we interpret the focus group results as preliminary complementary discussions with practitioners, avoiding validity claims. Nevertheless, we had rich discussions that nicely complemented our literature study findings. It is noteworthy that these discussions already revealed interesting facts, such as the potential unawareness of developers about the negative effects of GDEs, and some example arguments on the perceived pros and cons of including the GDEs in gamified software. Still, we are aware that a single focus group is not be enough to reach generalizable findings. Unfortunately, identifying teams with experience in developing gamified software and willing to collaborate with academia is not a trivial task. Therefore, conducting new focus group sessions is part of future work. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Clauvin Almeida, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil; (2) Marcos Kalinowski, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil; (3) Anderson Uchoa, Federal University of Ceara (UFC), Itapaje, Brazil; (4) Bruno Feijo, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil. Authors: Authors: (1) Clauvin Almeida, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil; (2) Marcos Kalinowski, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil; (3) Anderson Uchoa, Federal University of Ceara (UFC), Itapaje, Brazil; (4) Bruno Feijo, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rio de Janeiro, Brazil. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2. Background and Related Work and 2.1. Gamification 2. Background and Related Work and 2.1. Gamification 2.2. Game Design Elements and 2.3. Gamification Effects 2.2. Game Design Elements and 2.3. Gamification Effects 2.4. Related Work on Gamification Negative Effects 2.4. Related Work on Gamification Negative Effects 3. Systematic Mapping and 3.1. The Research Questions 3. Systematic Mapping and 3.1. The Research Questions 3.2. Search Strategy and 3.3. Inclusion and Exclusion Criteria 3.2. Search Strategy and 3.3. Inclusion and Exclusion Criteria 3.4. Applying the Search Strategy 3.4. Applying the Search Strategy 3.5. Data Extraction 3.5. Data Extraction 4. Systematic Mapping Results 4. Systematic Mapping Results 5. Focus Group: Developer Perception on the Negative Effects of Game Design Elements 5. Focus Group: Developer Perception on the Negative Effects of Game Design Elements 5.1. Context and Participant Characterization 5.1. Context and Participant Characterization 5.2. Focus Group Design 5.2. Focus Group Design 5.3. The Developers’ Perception on The Negative Effects 5.3. The Developers’ Perception on The Negative Effects 5.4. On the Perceived Usefulness, Ease of use and Intent of Adoption of Mapped Negative Effects 5.4. On the Perceived Usefulness, Ease of use and Intent of Adoption of Mapped Negative Effects 5.5. Participant Feedback 5.5. Participant Feedback 6. Limitations 6. Limitations 7. Concluding Remarks 7. Concluding Remarks 7.1. Future Research Directions 7.1. Future Research Directions Acknowledgements and References Acknowledgements and References 6. Limitations One of the limitations of our mapping study concerns potentially missing papers. After analyzing 3458 papers (see the list in our online repository), based on our inclusion and exclusion criteria, we initially included 79 papers. While our combined search strategy allowed identifying significantly more papers than the database search strategy employed by Toda et al. [27] (we identified 34 papers published between 2012 and 2016, against 17), the sets had differences. Therefore, we manually included the papers found by Toda et al., which were missed by our first search and extended searches, ending up with a final set of 87 papers. It is noteworthy that we verified that many of the missed papers would also have been retrieved by subsequent snowballing iterations. Nevertheless, an extension applying subsequent snowballing iterations and investigating different hybrid strategies [15], would require significant additional effort, involving analyzing several thousands of papers which could characterize a different publication instead of an extension. We are confident that our final set of included papers as part of this publication allowed providing an unbiased and meaningful overview of the adverse effects related to GDEs in gamified education software. Another risk of false negatives concerns the filtering process. We screened all papers considering only titles, abstracts, and keywords, which may not contain sufficient information to decide upon inclusion. We avoided applying EC1 and EC3 during the initial screening to lower this risk, only excluding papers that we had high confidence of not investigating GDE effects (EC1) and reporting negative effects (EC3). In case of any doubt, the paper was left for full text-based assessment. Moreover, the application of the inclusion and exclusion criteria made by the first author was reviewed by the second and fourth authors in meetings. In case of doubt, during the initial screening or full-text-based assessment, discussions were held to reach a consensus. Furthermore, we chose not to consider grey literature as part of our inclusion criteria. There is the possibility of relevant grey literature that does not have equivalent in non-grey academic papers. On the other hand, even though we did not explicitly evaluate the strength of evidence, the results herein reported are based on peer-reviewed research and backed by empirical studies. Moreover, while research that reaches negative results is essential because it shows us what does not work [115], there still seems to be a publication bias towards positive outcomes. Research reporting negative results tends to have less scientific interest, fewer citations, and to be less often published [116]. Hence, there may be additional negative results that were not published and which, for that reason, could not be included in our mapping study. Finally, in our summary of the negative effects of GDEs (cf. Figure 4), we included negative effects that may affect different roles (e.g., the students and the maintainers/teachers responsible for configuring the system and keeping it up and running with appropriate educational resources), as we found them reported in the literature. Our goal was to organize evidence regarding the overall adverse effects of GDEs in gamified education software. While we included an overview of the negative effects for different roles (e.g., RQ1.B and RQ1.C), we considered a fine-grained analysis of the effects of each GDEs for each of these roles out of the scope of our intended overview and part of future work. With regard to the focus group session, we wanted to gather insights on the mapped negative effects of game design elements from the point of view of practitioners. In a virtual focus group, generally, the number of participants is reduced, four being considered appropriate (our case) [117]. However, it is suggested to plan more online focus groups with fewer participants than when conducting face-to-face groups [118]. While we carefully planned our focus group, followed best practices, and carefully conducted our qualitative analyses, a single session is surely a limitation. Therefore, we interpret the focus group results as preliminary complementary discussions with practitioners, avoiding validity claims. Nevertheless, we had rich discussions that nicely complemented our literature study findings. It is noteworthy that these discussions already revealed interesting facts, such as the potential unawareness of developers about the negative effects of GDEs, and some example arguments on the perceived pros and cons of including the GDEs in gamified software. Still, we are aware that a single focus group is not be enough to reach generalizable findings. Unfortunately, identifying teams with experience in developing gamified software and willing to collaborate with academia is not a trivial task. Therefore, conducting new focus group sessions is part of future work. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

Limitations and Insights on The Effects of GDEs in Education

Authors:
(1) Dan Kondratyuk, Google Research and with Equal contribution;
(2) Lijun Yu, Google Research, Carnegie Mellon University and with Equal contribution;
(3) Xiuye Gu, Google Research and with Equal contribution;
(4) Jose Lezama, Google Research and with Equal contribution;
(5) Jonathan Huang, Google Research and with Equal contribution;
(6) Grant Schindler, Google Research;
(7) Rachel Hornung, Google Research;
(8) Vighnesh Birodkar, Google Research;
(9) Jimmy Yan, Google Research;
(10) Krishna Somandepalli, Google Research;
(11) Hassan Akbari, Google Research;
(12) Yair Alon, Google Research;
(13) Yong Cheng, Google DeepMind;
(14) Josh Dillon, Google Research;
(15) Agrim Gupta, Google Research;
(16) Meera Hahn, Google Research;
(17) Anja Hauth, Google Research;
(18) David Hendon, Google Research;
(19) Alonso Martinez, Google Research;
(20) David Minnen, Google Research;
(21) Mikhail Sirotenko, Google Research;
(22) Kihyuk Sohn, Google Research;
(23) Xuan Yang, Google Research;
(24) Hartwig Adam, Google Research;
(25) Ming-Hsuan Yang, Google Research;
(26) Irfan Essa, Google Research;
(27) Huisheng Wang, Google Research;
(28) David A. Ross, Google Research;
(29) Bryan Seybold, Google Research and with Equal contribution;
(30) Lu Jiang, Google Research and with Equal contribution. Table of Links Abstract and 1 Introduction 2. Related Work 3. Model Overview and 3.1. Tokenization 3.2. Language Model Backbone and 3.3. Super-Resolution 4. LLM Pretraining for Generation 4.1. Task Prompt Design 4.2. Training Strategy 5. Experiments 5.1. Experimental Setup 5.2. Pretraining Task Analysis 5.3. Comparison with the State-of-the-Art 5.4. LLM’s Diverse Capabilities in Video Generation and 5.5. Limitations 6. Conclusion, Acknowledgements, and References A. Appendix 5.4. LLM’s Diverse Capabilities in Video Generation This subsection presents several capabilities we discover from the pretrained VideoPoet, shedding light on the LLM’s promising potential in video generation. By combining the flexibility of our autoregressive language model to perform diverse tasks such as extending video in time, inpainting, outpainting, and stylization, VideoPoet accomplishes multiple tasks in a unified model. Coherent long video generation and image-to-video. A benefit of an decoder-based language model is that it pairs well with autoregressively extending generation in time. We present two different variants: Generating longer videos and converting images to videos. Encoding the first frame independently allows us to convert any image into the initial frame of a video without padding. Subsequent frames are generated by predicting remaining tokens, transforming the image into a video as shown in Fig. 6 [1]. based on previously generated video, and produces temporally consistent videos without significant distortion. Such capabilities are rarely observed in contemporary diffusion models. Zero-shot video editing and task chaining. With the multi-task pretraining, VideoPoet exhibits task generalization that can be chained together to perform novel tasks. We show the model can apply image-to-video animation followed by video-to-video stylization in Fig. 7. In the Appendix, Fig. 10 shows another example applying video-tovideo outpainting, followed by editing them with additional video-to-video effects. At each stage, the quality of the output is sufficient to remain in-distribution (i.e. teacher forcing) for the next stage without noticeable artifacts. These capabilities can be attributed to our multimodal task design within an LLM transformer framework that allows for modeling multimodal content using a single transformer architecture over a unified vocabulary. Zero-shot video stylization. Stylization results are presented in Appendix A.4 where the structure and text are used as prefixes to guide the language model. Unlike other stylization methods that employ adapter modules such as cross-attention networks (Zhang et al., 2023b) or latent blending (Meng et al., 2021), our approach stylizes videos within an LLM as one of several generative tasks. 3D structure, camera motion, visual styles. Even though we do not add specific training data or losses to encourage 3D consistency, our model can rotate around objects and predict reasonable visualizations of the backside of objects. Additionally, with only a small proportion of input videos and texts describing camera motion, our model can use short text prompts to apply a range of camera motions to image-to-video and text-to-video generations (see Fig. 11). 5.5. Limitations Despite VideoPoet demonstrating highly competitive performance of LLMs relative to state-of-the-art models, certain limitations are still observed. For example, the RGB frame reconstruction from compressed and quantized tokens place an upper bound on the generative model’s visual fidelity. Second, the per-frame aesthetic biases in static scenes does not match the best baseline. This difference is largely due to a choice of training data, where we focus our training on more natural aesthetics and excluded some sources containing copyrighted images, such as LAION (Schuhmann et al., 2022), which is commonly used in other work. Finally, small objects and fine-grained details, especially when coupled with significant motions, remains difficult within the token-based modeling. This paper is available on arxiv under CC BY 4.0 DEED license. [1] For image-to-video examples we source images from Wikimedia Commons: https://commons.wikimedia.org/wiki/Main Page Authors: (1) Dan Kondratyuk, Google Research and with Equal contribution; (2) Lijun Yu, Google Research, Carnegie Mellon University and with Equal contribution; (3) Xiuye Gu, Google Research and with Equal contribution; (4) Jose Lezama, Google Research and with Equal contribution; (5) Jonathan Huang, Google Research and with Equal contribution; (6) Grant Schindler, Google Research; (7) Rachel Hornung, Google Research; (8) Vighnesh Birodkar, Google Research; (9) Jimmy Yan, Google Research; (10) Krishna Somandepalli, Google Research; (11) Hassan Akbari, Google Research; (12) Yair Alon, Google Research; (13) Yong Cheng, Google DeepMind; (14) Josh Dillon, Google Research; (15) Agrim Gupta, Google Research; (16) Meera Hahn, Google Research; (17) Anja Hauth, Google Research; (18) David Hendon, Google Research; (19) Alonso Martinez, Google Research; (20) David Minnen, Google Research; (21) Mikhail Sirotenko, Google Research; (22) Kihyuk Sohn, Google Research; (23) Xuan Yang, Google Research; (24) Hartwig Adam, Google Research; (25) Ming-Hsuan Yang, Google Research; (26) Irfan Essa, Google Research; (27) Huisheng Wang, Google Research; (28) David A. Ross, Google Research; (29) Bryan Seybold, Google Research and with Equal contribution; (30) Lu Jiang, Google Research and with Equal contribution. Authors: Authors: (1) Dan Kondratyuk, Google Research and with Equal contribution; (2) Lijun Yu, Google Research, Carnegie Mellon University and with Equal contribution; (3) Xiuye Gu, Google Research and with Equal contribution; (4) Jose Lezama, Google Research and with Equal contribution; (5) Jonathan Huang, Google Research and with Equal contribution; (6) Grant Schindler, Google Research; (7) Rachel Hornung, Google Research; (8) Vighnesh Birodkar, Google Research; (9) Jimmy Yan, Google Research; (10) Krishna Somandepalli, Google Research; (11) Hassan Akbari, Google Research; (12) Yair Alon, Google Research; (13) Yong Cheng, Google DeepMind; (14) Josh Dillon, Google Research; (15) Agrim Gupta, Google Research; (16) Meera Hahn, Google Research; (17) Anja Hauth, Google Research; (18) David Hendon, Google Research; (19) Alonso Martinez, Google Research; (20) David Minnen, Google Research; (21) Mikhail Sirotenko, Google Research; (22) Kihyuk Sohn, Google Research; (23) Xuan Yang, Google Research; (24) Hartwig Adam, Google Research; (25) Ming-Hsuan Yang, Google Research; (26) Irfan Essa, Google Research; (27) Huisheng Wang, Google Research; (28) David A. Ross, Google Research; (29) Bryan Seybold, Google Research and with Equal contribution; (30) Lu Jiang, Google Research and with Equal contribution. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2. Related Work 2. Related Work 3. Model Overview and 3.1. Tokenization 3. Model Overview and 3.1. Tokenization 3.2. Language Model Backbone and 3.3. Super-Resolution 3.2. Language Model Backbone and 3.3. Super-Resolution 4. LLM Pretraining for Generation 4.1. Task Prompt Design 4.1. Task Prompt Design 4.2. Training Strategy 4.2. Training Strategy 5. Experiments 5.1. Experimental Setup 5.1. Experimental Setup 5.2. Pretraining Task Analysis 5.2. Pretraining Task Analysis 5.3. Comparison with the State-of-the-Art 5.3. Comparison with the State-of-the-Art 5.4. LLM’s Diverse Capabilities in Video Generation and 5.5. Limitations 5.4. LLM’s Diverse Capabilities in Video Generation and 5.5. Limitations 6. Conclusion, Acknowledgements, and References 6. Conclusion, Acknowledgements, and References A. Appendix A. Appendix 5.4. LLM’s Diverse Capabilities in Video Generation This subsection presents several capabilities we discover from the pretrained VideoPoet, shedding light on the LLM’s promising potential in video generation. By combining the flexibility of our autoregressive language model to perform diverse tasks such as extending video in time, inpainting, outpainting, and stylization, VideoPoet accomplishes multiple tasks in a unified model. Coherent long video generation and image-to-video. A benefit of an decoder-based language model is that it pairs well with autoregressively extending generation in time. We present two different variants: Generating longer videos and converting images to videos. Encoding the first frame independently allows us to convert any image into the initial frame of a video without padding. Subsequent frames are generated by predicting remaining tokens, transforming the image into a video as shown in Fig. 6 [1]. based on previously generated video, and produces temporally consistent videos without significant distortion. Such capabilities are rarely observed in contemporary diffusion models. Zero-shot video editing and task chaining. With the multi-task pretraining, VideoPoet exhibits task generalization that can be chained together to perform novel tasks. We show the model can apply image-to-video animation followed by video-to-video stylization in Fig. 7. In the Appendix, Fig. 10 shows another example applying video-tovideo outpainting, followed by editing them with additional video-to-video effects. At each stage, the quality of the output is sufficient to remain in-distribution (i.e. teacher forcing) for the next stage without noticeable artifacts. These capabilities can be attributed to our multimodal task design within an LLM transformer framework that allows for modeling multimodal content using a single transformer architecture over a unified vocabulary. Zero-shot video editing and task chaining. Zero-shot video stylization. Stylization results are presented in Appendix A.4 where the structure and text are used as prefixes to guide the language model. Unlike other stylization methods that employ adapter modules such as cross-attention networks (Zhang et al., 2023b) or latent blending (Meng et al., 2021), our approach stylizes videos within an LLM as one of several generative tasks. Stylization results are presented in Appendix A.4 where the structure 3D structure, camera motion, visual styles. Even though we do not add specific training data or losses to encourage 3D consistency, our model can rotate around objects and predict reasonable visualizations of the backside of objects. 3D structure, camera motion, visual styles. Additionally, with only a small proportion of input videos and texts describing camera motion, our model can use short text prompts to apply a range of camera motions to image-to-video and text-to-video generations (see Fig. 11). 5.5. Limitations Despite VideoPoet demonstrating highly competitive performance of LLMs relative to state-of-the-art models, certain limitations are still observed. For example, the RGB frame reconstruction from compressed and quantized tokens place an upper bound on the generative model’s visual fidelity. Second, the per-frame aesthetic biases in static scenes does not match the best baseline. This difference is largely due to a choice of training data, where we focus our training on more natural aesthetics and excluded some sources containing copyrighted images, such as LAION (Schuhmann et al., 2022), which is commonly used in other work. Finally, small objects and fine-grained details, especially when coupled with significant motions, remains difficult within the token-based modeling. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv [1] For image-to-video examples we source images from Wikimedia Commons: https://commons.wikimedia.org/wiki/Main Page

LLM’s Diverse Capabilities in Video Generation and Limitations

Table of Links Abstract and 1 Introduction 2. Related Work 3. Method and 3.1. Architecture 3.2. Loss and 3.3. Implementation Details 4. Data Curation 4.1. Training Dataset 4.2. Evaluation Benchmark 5. Experiments and 5.1. Metrics 5.2. Baselines 5.3. Comparison to SOTA Methods 5.4. Qualitative Results and 5.5. Ablation Study 6. Limitations and Discussion 7. Conclusion and References A. Additional Qualitative Comparison B. Inference on AI-generated Images C. Data Curation Details 6. Limitations and Discussion Due to computational resource limitations, we are not able to process and train our model on the full Objaverse dataset. Currently, the meshes from Objaverse we use only consist of 5% of Objaverse and 0.4% of Objaverse-XL objects. Based on the promising scaling properties of recent foundation models [12, 24, 61], we believe it will be valuable to explore the scaling properties of method. Another limitation of our work is that we have not considered the modeling of object texture. Predicting textures of unseen surfaces is highly ill-posed and can greatly benefit from a strong 2D prior. Given the recent success of 2D diffusion models [48] and their application in optimization-based 3D generation methods [7, 11, 29, 34, 40, 59], we think it will be promising to initialize or regularize these methods with our shape prior, potentially boosting both the optimization efficiency and generation quality. This paper is available on arxiv under CC BY 4.0 DEED license. Authors:
(1) Zixuan Huang, University of Illinois at Urbana-Champaign and both authors contributed equally to this work;
(2) Stefan Stojanov, Georgia Institute of Technology and both authors contributed equally to this work;
(3) Anh Thai, Georgia Institute of Technology;
(4) Varun Jampani, Stability AI;
(5) James M. Rehg, University of Illinois at Urbana-Champaign. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2. Related Work 2. Related Work 3. Method and 3.1. Architecture 3. Method and 3.1. Architecture 3.2. Loss and 3.3. Implementation Details 3.2. Loss and 3.3. Implementation Details 4. Data Curation 4.1. Training Dataset 4.1. Training Dataset 4.2. Evaluation Benchmark 4.2. Evaluation Benchmark 5. Experiments and 5.1. Metrics 5. Experiments and 5.1. Metrics 5.2. Baselines 5.2. Baselines 5.3. Comparison to SOTA Methods 5.3. Comparison to SOTA Methods 5.4. Qualitative Results and 5.5. Ablation Study 5.4. Qualitative Results and 5.5. Ablation Study 6. Limitations and Discussion 6. Limitations and Discussion 7. Conclusion and References 7. Conclusion and References A. Additional Qualitative Comparison A. Additional Qualitative Comparison B. Inference on AI-generated Images B. Inference on AI-generated Images C. Data Curation Details C. Data Curation Details 6. Limitations and Discussion Due to computational resource limitations, we are not able to process and train our model on the full Objaverse dataset. Currently, the meshes from Objaverse we use only consist of 5% of Objaverse and 0.4% of Objaverse-XL objects. Based on the promising scaling properties of recent foundation models [12, 24, 61], we believe it will be valuable to explore the scaling properties of method. Another limitation of our work is that we have not considered the modeling of object texture. Predicting textures of unseen surfaces is highly ill-posed and can greatly benefit from a strong 2D prior. Given the recent success of 2D diffusion models [48] and their application in optimization-based 3D generation methods [7, 11, 29, 34, 40, 59], we think it will be promising to initialize or regularize these methods with our shape prior, potentially boosting both the optimization efficiency and generation quality. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv Authors: (1) Zixuan Huang, University of Illinois at Urbana-Champaign and both authors contributed equally to this work; (2) Stefan Stojanov, Georgia Institute of Technology and both authors contributed equally to this work; (3) Anh Thai, Georgia Institute of Technology; (4) Varun Jampani, Stability AI; (5) James M. Rehg, University of Illinois at Urbana-Champaign. Authors: Authors: (1) Zixuan Huang, University of Illinois at Urbana-Champaign and both authors contributed equally to this work; (2) Stefan Stojanov, Georgia Institute of Technology and both authors contributed equally to this work; (3) Anh Thai, Georgia Institute of Technology; (4) Varun Jampani, Stability AI; (5) James M. Rehg, University of Illinois at Urbana-Champaign.

ZeroShape: The Limitations We Are Facing

This is Part 8 of a 11-part series based on the research paper “Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior”. Use the table of links below to navigate to the next part. Table of Links Part 1: Abstract & Introduction Part 2: Important Concepts Part 3: System Description Part 4: Agents & Simulation Details Part 5: Experiment Design Part 6: Continual Learning Part 7: Experiment Results Part 8: Market and Agent Responsiveness to External Events Part 9: Conclusion & References Part 10: Additional Simulation Results Part 11: Simulation Configuration 5.2 Market and Agent Responsiveness to External Events In the first experiment, we introduce a rule-based agent which performs a sequence of “flash sales”. During a flash sale event, this agent continuously sends sell orders (300 lots each) every second for a duration of 5 seconds. Then, the agent remains inactive for 400 seconds to allow the market to recover. The goal is to study how the market moves during and after the flash sale events and how the agents’ behavior changes after training with these events. Figure 7 depicts the price movement after the start of flash sale events. To create these plots we use the average price of multiple simulations normalized by the price at which flash sale started. The untrained group shows a significant and permanent price drop in contrast to the testing and continual training groups. In the testing group, prices recover to pre-flash-sale levels. However, the continual training group shows more realistic trajectories, with prices dropping and recovering to levels lower than the pre-sale prices. This price pattern aligns with descriptions of permanent and temporary price impact in [24]. We are also interested in how RL agents’ policy functions are affected when trained with flash sale events. Recall the action symmetric tweak and asymmetric tweak parameters control the prices of the limit orders sent by MM’s (see (1) and illustration in Figure 1). Intuitively, a larger symmetric tweak makes the agent place orders away from the best bid and ask (more conservatively). A positive asymmetric tweak indicates that the MM expects future higher prices than the current one, while a negative value indicates the MM expects the market to go down. When a flash sale event happens, market sell orders take liquidity from the bid side of the LOB and this results in an imbalanced LOB. In contrast, a balanced LOB means the volumes on the buy and sell sides are comparable. To determine the MM agents’ reaction we collect the states with balanced and imbalanced LOBs, and feed these states to the different agents’ policy functions. Figure 8 shows the action symmetric tweak values in the top row and the asymmetric tweak in the bottom row. We can see that when the order book is balanced the three groups output a similar distribution of actions (Figure 8b and 8d). However, when fed imbalanced LOBs (Figures 8a and 8c), the agents in the continual training group output higher symmetric price tweaks and lower asymmetric price tweaks, compared with the agents in the two other groups. Thus when a flash sale event is detected, the MM agents in the continual learning group tend to place orders deeper into the LOB and move the expected price equilibrium downward. This behavior aligns with the findings in literature [25, 26, 27]. These results show that the agents can adapt to different market conditions through continual learning. In this experiment, we only use continual learning Group A MM agents, as they need to adapt to changing market conditions. The LT agents are also from Group A but their reward function is changed through the evolution of the target buy/sell parameters. Specifically, Figure 9 shows the price process resulting from the activity of these informed LT agents. The four phases separated by dashed red lines are: Sell (0.3/0.4), Buy (0.4/0.35), Balanced buy and sell (0.4/0.4), and last Buy (0.4/0.3), the numbers in parenthesis indicate (buy fraction/sell fraction). We expect to see the price movement aligning with the target buy/sell parameters. Additionally, similar to the previous study we collect states in the first and the last phases (i.e., steps 0-10,000 when the price goes down and steps 30,000-36,000 when the price goes up). We feed these states to the MM’s policy function at the beginning of the day (Before) and at the end of the trading day (After). Figure 10 shows the distributions of the outputs in action symmetric and asymmetric tweak. In both scenarios when the price is going down and when the price is going up, the action symmetric tweak (top row) gets larger, thus the MM agent becomes more conservative and tends to enlarge the spread. This aligns with the findings in the previous experiment. We can also see that the distribution of price asymmetric tweak becomes negative when prices are going down and becomes positive when prices are going up. This means the learned agents change their expectations of future prices along with the observed market direction. In contrast, the asymmetric tweak average is close to zero for agents that are not continuously trained. This helps explain why the price tends to come back to pre-flash-sale values for the agents in Group B in Figure 7. Authors:
(1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu);
(2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu);
(3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu);
(4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This is Part 8 of a 11-part series based on the research paper “Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior”. Use the table of links below to navigate to the next part. This is Part 8 of a 11-part series based on the research paper “ Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior ”. Use the table of links below to navigate to the next part. Part 8 Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior Table of Links Part 1: Abstract & Introduction Part 1: Abstract & Introduction Part 2: Important Concepts Part 2: Important Concepts Part 3: System Description Part 3: System Description Part 4: Agents & Simulation Details Part 4: Agents & Simulation Details Part 5: Experiment Design Part 5: Experiment Design Part 6: Continual Learning Part 6: Continual Learning Part 7: Experiment Results Part 7: Experiment Results Part 8: Market and Agent Responsiveness to External Events Part 8: Market and Agent Responsiveness to External Events Part 9: Conclusion & References Part 9: Conclusion & References Part 10: Additional Simulation Results Part 10: Additional Simulation Results Part 11: Simulation Configuration Part 11: Simulation Configuration 5.2 Market and Agent Responsiveness to External Events In the first experiment, we introduce a rule-based agent which performs a sequence of “flash sales”. During a flash sale event, this agent continuously sends sell orders (300 lots each) every second for a duration of 5 seconds. Then, the agent remains inactive for 400 seconds to allow the market to recover. The goal is to study how the market moves during and after the flash sale events and how the agents’ behavior changes after training with these events. Figure 7 depicts the price movement after the start of flash sale events. To create these plots we use the average price of multiple simulations normalized by the price at which flash sale started. The untrained group shows a significant and permanent price drop in contrast to the testing and continual training groups. In the testing group, prices recover to pre-flash-sale levels. However, the continual training group shows more realistic trajectories, with prices dropping and recovering to levels lower than the pre-sale prices. This price pattern aligns with descriptions of permanent and temporary price impact in [24]. We are also interested in how RL agents’ policy functions are affected when trained with flash sale events. Recall the action symmetric tweak and asymmetric tweak parameters control the prices of the limit orders sent by MM’s (see (1) and illustration in Figure 1). Intuitively, a larger symmetric tweak makes the agent place orders away from the best bid and ask (more conservatively). A positive asymmetric tweak indicates that the MM expects future higher prices than the current one, while a negative value indicates the MM expects the market to go down. When a flash sale event happens, market sell orders take liquidity from the bid side of the LOB and this results in an imbalanced LOB. In contrast, a balanced LOB means the volumes on the buy and sell sides are comparable. To determine the MM agents’ reaction we collect the states with balanced and imbalanced LOBs, and feed these states to the different agents’ policy functions. Figure 8 shows the action symmetric tweak values in the top row and the asymmetric tweak in the bottom row. We can see that when the order book is balanced the three groups output a similar distribution of actions (Figure 8b and 8d). However, when fed imbalanced LOBs (Figures 8a and 8c), the agents in the continual training group output higher symmetric price tweaks and lower asymmetric price tweaks, compared with the agents in the two other groups. Thus when a flash sale event is detected, the MM agents in the continual learning group tend to place orders deeper into the LOB and move the expected price equilibrium downward. This behavior aligns with the findings in literature [25, 26, 27]. These results show that the agents can adapt to different market conditions through continual learning. In this experiment, we only use continual learning Group A MM agents, as they need to adapt to changing market conditions. The LT agents are also from Group A but their reward function is changed through the evolution of the target buy/sell parameters. Specifically, Figure 9 shows the price process resulting from the activity of these informed LT agents. The four phases separated by dashed red lines are: Sell (0.3/0.4), Buy (0.4/0.35), Balanced buy and sell (0.4/0.4), and last Buy (0.4/0.3), the numbers in parenthesis indicate (buy fraction/sell fraction). We expect to see the price movement aligning with the target buy/sell parameters. Additionally, similar to the previous study we collect states in the first and the last phases (i.e., steps 0-10,000 when the price goes down and steps 30,000-36,000 when the price goes up). We feed these states to the MM’s policy function at the beginning of the day (Before) and at the end of the trading day (After). Figure 10 shows the distributions of the outputs in action symmetric and asymmetric tweak. In both scenarios when the price is going down and when the price is going up, the action symmetric tweak (top row) gets larger, thus the MM agent becomes more conservative and tends to enlarge the spread. This aligns with the findings in the previous experiment. We can also see that the distribution of price asymmetric tweak becomes negative when prices are going down and becomes positive when prices are going up. This means the learned agents change their expectations of future prices along with the observed market direction. In contrast, the asymmetric tweak average is close to zero for agents that are not continuously trained. This helps explain why the price tends to come back to pre-flash-sale values for the agents in Group B in Figure 7. Authors: (1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu); (2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu); (3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu); (4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). Authors: Authors: (1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu); (2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu); (3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu); (4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv

RL Agents Adapt to Flash Sale Events & Imbalanced Limit Order Books (LOBs)

LLaVA-Phi: Limitations and What You Can Expect in the Future

Stories, News & Mentions about LIMIT

Limit’s Stories on HackerNoon