⚒️ Can LMs (especially reasoning models) effectively self-refine their responses when prompted to do so? In our new challenging benchmark, RefineBench, we revisit this question and show that the answer is still "no"-but there is a nuance! 🤗 huggingface.co/papers/2511.2…
MACHINE LEARNING
-
Context Engineering Clearly Explained with Visuals
By
–
Context engineering, clearly explained (with visuals) #AI @mvollmer1 @enilev @anijov @CatherineAdenle @FmFrancoise @AkwyZ @avrohomg @jblefevre60 @BetaMoroney @gvalan #MachineLearning @CurieuxExplorer @BIScorecard @Nicochan33 @Kevin_ODonovan @Hana_ElSayyed @sallyeaves
→ View original post on X — @marcusborba, 2025-12-01 02:59 UTC
-
Industrial AI Data Quality and Preprocessing for Manufacturing
By
–
Your AI is only as smart as your data is clean.
— Lucian Fogoros (@fogoros) 30 novembre 2025
Erikjan Franssen of @adlibsoftware & Cesar Bravo of @honeywell explain how preprocessing & context keep industrial AI from learning the wrong lessons. https://t.co/u8mjZDuZIg #sponsored #adlib_iiot #AI #Manufacturing @IIoT_World pic.twitter.com/y9AjDlzxukYour AI is only as smart as your data is clean. Erikjan Franssen of @adlibsoftware & Cesar Bravo of @honeywell explain how preprocessing & context keep industrial AI from learning the wrong lessons. https://
buff.ly/7ge1Hbq #sponsored #adlib_iiot #AI #Manufacturing @IIoT_World -
Gemini vs ChatGPT vs Claude: Heptagon Physics Demo
By
–
Gemini 3 Pro (left) vs ChatGPT 5.1 thinking vs Claude Opus 4.5 (right)
— God of Prompt (@godofprompt) 30 novembre 2025
Prompt:
Multi-ball heptagon physics
Write html, css, and javascript to display 20 balls bouncing inside a spinning heptagon. each ball should have a unique color, a number label, gravity, and collision… pic.twitter.com/YvZ7jC8Mc1Gemini 3 Pro (left) vs ChatGPT 5.1 thinking vs Claude Opus 4.5 (right) Prompt:
Multi-ball heptagon physics
Write html, css, and javascript to display 20 balls bouncing inside a spinning heptagon. each ball should have a unique color, a number label, gravity, and collision -
Shorter reasoning paths outperform longer ones in AI models
By
–
Ever wondered why large reasoning models sometimes overcomplicate problems? This study finds shorter reasoning paths consistently outperform longer ones across stochastic decodes, but exhaustive exploration of the tree-like reasoning space is impossible due to exponential
-
Memory-Efficient Embodied Agents Training with Reinforcement Learning
By
–
Memo: Training Memory-Efficient Embodied Agents with Reinforcement Learning Gupta et al.: https://
arxiv.org/abs/2510.19732 #ArtificialIntelligence #DeepLearning #MachineLearning -
DeepSeek-Math-V2 Reaches Gold Medal Performance on Putnam
By
–
2. DeepSeek’s math model reaches gold-medal performance DeepSeek-Math-V2, an open-source MoE model, scored 118/120 on the 2024 Putnam (above top human competitors) and hit gold-level performance on IMO 2025 problems. Open-source reasoning keeps getting scarier.
-
Major Neural Networks Overview Chart
By
–
All major neural networks, in one chart. v/The Asimov Institute
-
Cognitive Foundations for Reasoning in Large Language Models
By
–
10. Cognitive Foundations for Reasoning in LLMs Develops a taxonomy of 28 cognitive elements and evaluate 192K reasoning traces from 18 models, plus human think-aloud traces.
-
Multi-Agent Collaboration Enables Multimodal LLM Vision Integration
By
–
9. Multi-Agent Collaboration for Multimodal LLMs introduces a framework where vision models serve as “eyes” for language models through multi-agent collaboration.