Hacker News
- Reasoning models are just LLMs https://antirez.com/news/146 14 comments
- A Visual Guide to Reasoning LLMs https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-reasoning-llms 2 comments
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL https://arxiv.org/abs/2501.12948 1056 comments
- LLMs don't do formal reasoning https://garymarcus.substack.com/p/llms-dont-do-formal-reasoning-and 121 comments
- Understanding the Limitations of Mathematical Reasoning in LLMs https://arxiv.org/abs/2410.05229 266 comments
- LLMs and Diagnostic Reasoning: A Randomized Clinical Vignette Study [pdf] https://www.medrxiv.org/content/10.1101/2024.03.12.24303785v1.full.pdf 3 comments
- Inductive or deductive? Rethinking the fundamental reasoning abilities of LLMs https://arxiv.org/abs/2408.00114 169 comments
- Simple tasks showing reasoning breakdown in state-of-the-art LLMs https://arxiv.org/abs/2406.02061 380 comments
- LLMs approach expert-level clinical knowledge and reasoning in ophthalmology https://www.ft.com/content/5b7a76be-467c-4074-8fd0-3e297bcd91d7 102 comments
- Yann LeCun and Geoffrey Hinton disagree whether LLMs can reason https://i.imgur.com/yjhL56T.png 6 comments
- LLMs cannot find reasoning errors, but can correct them https://arxiv.org/abs/2311.08516 142 comments
- Thought Propagation: An analogical approach to complex reasoning with LLMs https://paperswithcode.com/paper/thought-propagation-an-analogical-approach-to 13 comments
- LLMs can't self-correct in reasoning tasks, DeepMind study finds https://bdtechtalks.com/2023/10/09/llm-self-correction-reasoning-failures/ 358 comments
- Can LLMs Reason and Plan? https://cacm.acm.org/blogs/blog-cacm/276268-can-llms-really-reason-and-plan/fulltext 46 comments
- Engineering Reasoning LLMs: Notes and Observations https://www.thelis.org/blog/reasoning-model-notes 0 comments
- Show HN: LLMs Playing Mafia games – See them lie, deceive, and reason https://mafia.opennumbers.xyz/ 22 comments
- LLMs can't perform "genuine logical reasoning," Apple researchers suggest https://arstechnica.com/ai/2024/10/llms-cant-perform-genuine-logical-reasoning-apple-researchers-suggest/ 73 comments
- COSP and USP: New methods to advance reasoning in LLMs https://pub.towardsai.net/inside-cosp-and-usp-google-research-new-methods-to-advance-reasoning-in-llms-07338b323dfd 4 comments
- Chain-of-Thought Hub: Measuring LLMs' Reasoning Performance https://github.com/FranxYao/chain-of-thought-hub 26 comments
- Numerical Precision Affects Mathematical Reasoning Capabilities of LLMs https://arxiv.org/abs/2410.13857 49 comments
- Q*: Improving Multi-Step Reasoning for LLMs with Deliberative Planning https://arxiv.org/abs/2406.14283 3 comments
Lobsters
- Reasoning models are just LLMs https://antirez.com/news/146 15 comments ai
- LLMs don’t do formal reasoning - and that is a HUGE problem https://garymarcus.substack.com/p/llms-dont-do-formal-reasoning-and 66 comments ai
- "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning", Guo et al 2025 {DeepSeek} https://arxiv.org/abs/2501.12948#deepseek 2 comments reinforcementlearning
- [R] Towards Time Series Reasoning with LLMs https://arxiv.org/abs/2409.11376 2 comments machinelearning
- Apple study exposes deep cracks in LLMs’ “reasoning” capabilities https://arstechnica.com/ai/2024/10/llms-cant-perform-genuine-logical-reasoning-apple-researchers-suggest/ 89 comments futurology
- DeepMind framework offers breakthrough in LLMs’ reasoning https://www.artificialintelligence-news.com/2024/02/08/deepmind-framework-offers-breakthrough-llm-reasoning/ 3 comments artificial
- Teaching LLMs to be more reasonable https://arxiv.org/abs/2210.07128 3 comments artificial
- [D] "Knowledge" vs "Reasoning" in LLMs https://arxiv.org/pdf/2206.04615.pdf 30 comments machinelearning
- Can LLMs really reason and plan? [D] https://medium.com/aiguys/can-llms-really-reason-and-plan-50b0ac6addd8 2 comments machinelearningnews
- Can LLMs Really Reason and Plan? https://cacm.acm.org/blogs/blog-cacm/276268-can-llms-really-reason-and-plan/fulltext 5 comments machinelearningnews
- How Google DeepMind's AlphaGeometry Reached Math Olympiad Level Reasoning By Combining Creative LLMs With Deductive Symbolic Engines: A visual guide https://codecompass00.substack.com/p/google-deepmind-alpha-geometry-neuro-symbolic-llm-system 2 comments deeplearning
- How Google DeepMind's AlphaGeometry Reached Math Olympiad Level Reasoning By Combining Creative LLMs With Deductive Symbolic Engines https://codecompass00.substack.com/p/google-deepmind-alpha-geometry-neuro-symbolic-llm-system 16 comments programming
- [R] rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking https://arxiv.org/abs/2501.04519 28 comments machinelearning
- Large Language Models and the Socratic Method: Exploring how LLMs can simulate Socratic dialogues to stimulate critical thinking. Introducing the Tree of Thoughts method to improve AI’s performance on complex reasoning tasks and emphasizing the importance of critical thinking in the era of AI. https://www.cbrincoveanu.com/posts/large-language-models-and-the-socratic-method/ 2 comments learnmachinelearning
- The AI Bubble may be about to burst. LLMs have reached the point of diminishing returns, and there's no sign of scaling leading to independent reasoning, needed for the first steps to AGI. https://garymarcus.substack.com/p/confirmed-llms-have-indeed-reached 285 comments futurology
- Largest text-to-speech AI model yet shows 'emergent abilities' | For reasons unknown to us, once LLMs grow past a certain point, they start being able to perform tasks they weren’t trained to do https://techcrunch.com/2024/02/14/largest-text-to-speech-ai-model-yet-shows-emergent-abilities/ 23 comments technews
- Chain together LLMs for reasoning and orchestrate multiple large models for accomplishing complex tasks like phoning someone using a GPT-4 model https://github.com/jina-ai/agentchain 3 comments python
- I've simulated hundreds of Mafia games where LLMs are the players - See them lie, deceive, and reason in real-time https://mafia.opennumbers.xyz 39 comments programming
- New CSAIL research highlights how LLMs excel in familiar scenarios but struggle in novel ones, questioning their true reasoning abilities versus reliance on memorization. https://news.mit.edu/2024/reasoning-skills-large-language-models-often-overestimated-0711 2 comments deeplearning