Live and recent posts from across the Fediverse tagged #ai-research, aggregated by home.social.
🌱 From RC Trust intern to PhD researcher.
During her RC Trust Internship, Nadya Hanaveriesa explored strategic reasoning in human-AI interaction with the HU[AM] group.
Several models appeared to misjudge how deeply humans would reason. In December, Nadya will return to Nils Köbis’ group as a PhD researcher.
🔗 https://rc-trust.ai/news/news-detail/from-rc-trust-intern-to-phd-researcher
Researchers have built JEPA-Anything, a framework that uses one shared recipe to create world models across seven different fields including vision, biology, control, molecular dynamics and weather. It splits a model predator into four orthogonal factors. https://www.marktechpost.com/2026/10/05/beyond-domain-specific-world-models-jepa-anything-uses-1-recipe-for-7-fields/ #AIagent #AI #GenAI #AIResearch
🚀 Fastest-growing AI projects today
1. The growth of these repositories reflects the ongoing trend towards more comprehensive...
2. **amitshekhariitbhu/ai-system-design**: Threpository a step-by-step guide on designing...
3. With a growth score of 54.21 and over 500 stars, it's clear that developers increasingl...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-october-02-2026
Wikimedia says OpenAI agents may have contributed to partial outage: Bots account for 65% of Wikimedia's most resource-heavy traffic, leaving its volunteers to clean up the mess. Will AI operators make agents easy to identify? https://ppc.land/wikimedia-says-openai-agents-may-have-contributed-to-partial-outage/ #Wikimedia #OpenAI #AIResearch #Bots #TechNews
🚀 Fastest-growing AI projects today
1. The repository with the highest growth score "ai-system-design" by amitshekhariitbhu, w...
2. **amitshekhariitbhu/ai-system-design**: Thproject a step-by-step guide to building AI s...
3. Its growth score of 56.42 and steady increase in stars indicate growing interest among...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-october-01-2026
Seventeen Countries Back a Kyoto Vision for AI-Powered Science
A declaration in the city of discovery Picture a laboratory where an AI system scans thousands of papers, proposes a promising experiment, and helps researchers decide what to test next. Now picture that laboratory connected to s…
#Japan #JP #Kyoto #AIinScience #AIResearch #artificialintelligence #kyotonews #MissionGenesis #news #京都 #京都府
https://www.alojapan.com/1549095/seventeen-countries-back-a-kyoto-vision-for-ai-powered-science/
https://www.alojapan.com/1549095/seventeen-countries-back-a-kyoto-vision-for-ai-powered-science/ Seventeen Countries Back a Kyoto Vision for AI-Powered Science #AIInScience #AIResearch #ArtificialIntelligence #Japan #Kyoto #KyotoNews #MissionGenesis #news #京都 #京都府 A declaration in the city of discovery Picture a laboratory where an AI system scans thousands of papers, proposes a promising experiment, and helps researchers decide what to test next. Now picture that laboratory connected to scientific teams across several countries. That is
Alibaba's Qwen AI has grown from a 7B parameter model in 2023 to a 2.4-trillion-parameter open-weight model by 2026. The journey spans dozens of releases, from early chatbots to multimodal models and reasoning systems. https://www.marktechpost.com/2026/10/04/the-story-of-qwen-alibabas-ai-models-from-7b-to-2-4t/ #AIagent #AI #GenAI #AIResearch
Alibaba's Qwen AI has grown from a 7B parameter model in 2023 to a 2.4-trillion-parameter open-weight model by 2026. The journey spans dozens of releases, from early chatbots to multimodal models and reasoning systems. https://www.marktechpost.com/2026/10/04/the-story-of-qwen-alibabas-ai-models-from-7b-to-2-4t/ #AIagent #AI #GenAI #AIResearch
Seventeen Countries Back a Kyoto Vision for AI-Powered Science
A declaration in the city of discovery Picture a laboratory where an AI system scans thousands of papers,…
#EuropeSays #Japan #JP #Kyoto #AIinScience #AIResearch #artificialintelligence #MissionGenesis
https://www.europesays.com/japan/101682/
🚀 Fastest-growing AI projects today
1. Among the most notable projects thweek `amitshekhariitbhu/ai-system-design`, which an e...
2. `amitshekhariitbhu/ai-system-design` a comprehensive guide to building AI systems lever...
3. With its recent spike in popularity—highlighted by a growth score of 66.60 and over 468...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-30-2026
Step 7 of "Proof Chain for the 24000/10451 Zero-Density Bound (Friendly Version)" (https://doi.org/10.13140/RG.2.2.11136.60161) is the key to achieve a very small, yet interesting improvement in the 30/13 Zero-Density Bound of Guth-Maynard. And I will never tire of saying it: what matters here is not the numerical improvement itself, but the technique used to achieve it.
This approach can be extrapolated to other mathematical problems that reflect onto physics, engineering, medicine, biology, and so on. In certain fields, even a minor numerical improvement can mean the difference between success and failure. Furthermore, this technique could be a game-changer for computing systems performing calculations similar to those used by Guth-Maynard, especially if they are recursive, where errors can propagate or accumulate, or where the full potential of the computation is simply not being maximized.
If proven effective in improving something as demanding as calculating the 30/13 bound, we will be looking at a highly useful tool for the entire scientific community. I am enthusiastically working to verify this work in Lean, but my resources are ridiculously limited. If anyone with computational power is willing to help, please contact me so we can team up. The verification is progressing well (considering Lean's limitations within this specific scope...), but it is slow. It is a pity, given the magnitude of this achievement and how much could be learned from this new tool.
🚀 Fastest-growing AI projects today
1. One standout project focusing on AI system design has seen significant growth, while ot...
2. **amitshekhariitbhu/ai-system-design**: Threpository aims to teach the step-by-step pro...
3. With a growth score of 82.00 and over 461 stars, it demonstrates strong interest from d...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-29-2026
Aleph Alpha has released Kolibri, an open-weight mixture-of-experts model with 78.1 billion parameters but only 3.46 billion active parameters. The German AI company's model supports both English and German languages, optimised for efficient inference. https://www.marktechpost.com/2026/10/04/aleph-alpha-releases-kolibri-a-78-1b-open-weight-english-german-moe-model-with-only-3-46b-active-parameters/ #AIagent #AI #GenAI #AIResearch
🚀 Fastest-growing AI projects today
1. Projects focusing on areas like multimodal models, video-related AI skills, and game li...
2. OmniJev's "awesome-jev-gallery" a repository that collects papers, open reproductions,...
3. With a growth score of 40.64 and over 465 stars, the project stands out for its compreh...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-28-2026
🚀 Fastest-growing AI projects today
1. Among these, one repository stands out for its comprehensive approach to reproducing an...
2. **OmniJev/awesome-jev-gallery**: Threpository compiles papers, open reproductions, and...
3. With its high growth score of 42.10 and 449 stars, it appears to be gaining significant...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-27-2026
Do you wonder how kinky LLMs are?
Well say no more, somebody did the research
The mathematics of the future will benefit from out of the box thinking, which is sometimes not closely aligned with academia. A bug left my old phone with half a million contacts for years. However, I did not fix it because I discovered that apps trying to read the address book would crash. I had just discovered a wonderful protection because even though I granted granular access to my contacts, if an app attempted to access the entire address book, it simply could not due to architecture and resource limitations.
The academic approach would have been to delete duplicate contacts. Instead, I preferred to observe the app ecosystem and see how they crashed when trying to access the address book, even when they supposedly did not have permission. This allowed me to understand that some apps were somehow bypassing that restriction, which a later debug confirmed. With an academic approach, I would have missed out on that invaluable experiment.
I believe that mathematics, with AI, is entering a fascinating new dimension where both worlds, the academic one and that of simple intuition, will finally unite. What awaits us far exceeds the most fertile imagination. The lines between right and wrong, and the procedures for researching and confirming advances, are going to bring many surprises. We are on the edge of an event horizon in the cognitive realm.
🚀 Fastest-growing AI projects today
1. One standout project OmniJev/awesome-jev-gallery, which curates papers and reproduction...
2. OmniJev's awesome-jev-gallery compiles a collection of resources including papers, open...
3. Its rapid rise in popularity highlighted by a robust growth score of 42.11 and nearly 3...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-26-2026
Decision AI models are emerging as a new category that returns calibrated probabilities instead of generated text. TypeSafe Jev and Fastino GLiDE represent this shift, offering responses that code can directly branch on. The approach promises faster response times and clearer confidence levels for enterprise applications. https://www.marktechpost.com/2026/10/02/decision-ai-models-explained-typesafe-jev-vs-fastino-glide-gliner2-5-decide-and-open-source-competitors/ #AIagent #AI #GenAI #AIResearch
Understanding Frontier Artificial Intelligence
https://casp.ac/reports/intelligence-explosion
Comments: https://news.ycombinator.com/item?id=49942002
#HackerNews #FrontierAI #ArtificialIntelligence #TechTrends #AIResearch #FutureOfAI
🚀 Fastest-growing AI projects today
1. The community's focus on practical applications and reproducibility evident in these gr...
2. **eternityspring/reelbench-skills** tracks learning notes and tooling skills related to...
3. With a growth score of 44.61 and over 835 stars, threpository attracting attention due...
Full report → https://pullrepo.com/report/todays-ai-research-fastest-growing-projects-september-25-2026
Researchers have identified thousands of new tells that distinguish AI-generated writing from human text, moving beyond outdated markers like em dashes. The study reveals how modern AI models leave distinct patterns that can be detected through linguistic analysis. https://gizmodo.com/this-matters-researchers-identify-thousands-of-new-tells-in-ai-writing-2000820447 #AIResearch #AI #GenAI
Researchers have built an AI that finally beat the world's best Stratego player. Ataraxos, from Carnegie Mellon, MIT, NYU and Stanford, defeated Pim Niemeijer 15-1 with four draws - using just 16 GPUs. The game has more than a decillion possible setups. https://arstechnica.com/science/2026/10/ai-finally-beat-the-best-stratego-player-in-history-and-did-it-on-a-budget/ #AIagent #AI #GenAI #AIResearch
👀
PewDiePie browsing cs.CR?
Video source: https://www.youtube.com/watch?v=ODDJXGY_1kQ
Article in the screenshot:
https://arxiv.org/abs/2608.09867
#cybersecurity #infosec #AI #AIResearch #AISecurity #Pewdiepie
Nvidia has released Kumo, a tabular AI model that outperforms GPT-4o on table understanding tasks. The model is available for research and enterprise use, representing Nvidia's expansion into AI foundation models. https://www.marktechpost.com/2026/09/30/nvidia-releases-kumo-tabular/ #AIagent #AI #GenAI #AIResearch
https://www.europesays.com/people/249946/ The ‘Godmother of AI’ Helped Spark Nvidia’s AI Boom. Now She’s Joining AMD. #AdvancedMicroDevices #AIResearch #AMDStock #ImageNet #LisaSu #Nvidia #WorldLabs
I am impressed by how new AI mastering of hidden information can lead to stronger, faster, and cheaper decision making with real world impact from negotiations to cybersecurity. This shows how imperfect information can still drive smart AI progress.
RT @OrcaRouter: Anthropic hat gerade den Alarm für die Cyber-Fähigkeiten von GLM-5.3 ausgelöst. Wir haben darauf aufgebaut. OrcaCyber Zero 1.0 basiert auf GLM-5.3 und wurde speziell für Schwachstellenforschung, Reproduktion von Exploits, Red-Teaming und autonome Cyber-Arbeitsabläufe nachtrainiert. 98,01 % Best-Of-1-Rate auf CyberGym L1. https://t.co/oXUHCB1OZU
mehr auf Arint.info
#AIResearch #CyberSecurity #GLM53 #OrcaCyber #RedTeaming #VulnerabilityResearch #arint_info
AI-Weekly for Tuesday, September 29, 2026 - Issue 236 | By Aaron Di Blasi, Publisher | Courtesy of the PWD Media Co-Op
https://ai-weekly.ai/newsletter-09-29-2026/
✨ The Week's News in Artificial Intelligence
A Mind Vault Solutions, Ltd. Publication
#ai #news #AINews #ArtificialIntelligence #AIWeekly #technology #tech #TechNews #TechTrends #MachineLearning #robotics #DataScience #AIResearch #FutureTech
Email Subscribers: 53,689 🔢️
Social Media: 199,069 🔢️
The world's #1 resource for AI news and trends.
AI-Weekly for Tuesday, September 29, 2026 - Issue 236 | By Aaron Di Blasi, Publisher | Courtesy of the PWD Media Co-Op
https://ai-weekly.ai/newsletter-09-29-2026/
✨ The Week's News in Artificial Intelligence
A Mind Vault Solutions, Ltd. Publication
#ai #news #AINews #ArtificialIntelligence #AIWeekly #technology #tech #TechNews #TechTrends #MachineLearning #robotics #DataScience #AIResearch #FutureTech
Email Subscribers: 53,689 🔢️
Social Media: 199,069 🔢️
The world's #1 resource for AI news and trends.
AI-Weekly for Tuesday, September 29, 2026 - Issue 236 | By Aaron Di Blasi, Publisher | Courtesy of the PWD Media Co-Op
https://ai-weekly.ai/newsletter-09-29-2026/
✨ The Week's News in Artificial Intelligence
A Mind Vault Solutions, Ltd. Publication
#ai #news #AINews #ArtificialIntelligence #AIWeekly #technology #tech #TechNews #TechTrends #MachineLearning #robotics #DataScience #AIResearch #FutureTech
Email Subscribers: 53,689 🔢️
Social Media: 199,069 🔢️
The world's #1 resource for AI news and trends.
Google Research has open-sourced RRSI, a method that enables AI agents to improve their own training harness without overfitting. The approach allows agents to adapt their learning environment dynamically, potentially leading to more robust autonomous systems. https://www.marktechpost.com/2026/09/29/google-research-open-sources-rrsi-ai-agents-that-improve-their-own-harness-without-overfitting/ #AIagent #AI #GenAI #AIResearch
MIT researchers have used an AI algorithm to develop a new formulation for RNA vaccines that makes them more heat-resistant. The vaccines could remain stable at room temperature for up to a year, or at nearly 100F for two months. This could help distribute vaccines to regions lacking cold-storage facilities. https://news.mit.edu/2026/new-formulation-helps-rna-vaccines-withstand-high-temperatures-0928 #AIagent #AI #GenAI #AIResearch
Fireworks AI has launched Ember-1, a post-trained version of Kimi K3 that produces 40% fewer tokens while maintaining quality. The model trims reasoning traces to cut costs for multi-step agentic workloads. Available as a research preview via Fireworks API. https://www.marktechpost.com/2026/09/28/fireworks-ai-releases-ember-1-a-post-trained-kimi-k3-that-uses-about-40-fewer-tokens/ #AIInfrastructure #AI #GenAI #AIResearch
TabPFN and TabICL vs. tuned XGBoost: the model that doesn't train won 14/14
https://efraingaray.com/en/blog/tabpfn-vs-xgboost/
Comments: https://news.ycombinator.com/item?id=49872864
#HackerNews #TabPFN #TabICL #XGBoost #machinelearning #AIresearch
This video is fantastic. I would recommend anyone interested in AI use and AI safety watch it.
To me, it seems like the dire predictions are more about marketing than science.
