Knowledge catalogue
Sub-32B open weights models now offer GPT-5 level intelligence with Qwen3.5 27B (Reasoning) matching GPT-5 (medium) at 42 and Gemma 4 31B (Reasoning) matching GPT-5 (low) at 39 on the Artificial Analy
Simply retrieving a reasoning trace looks a lot like human reasoning, until it's time to navigate uncharted territory. If you memorized all reasoning traces of humans from 10,000 BC, you could automat
Interesting paper to improve recurrent reasoning. Looped models are great because you get more reasoning out of a model without adding parameters. So this work proposes a looped architecture with a ne
Very interesting paper on LLM reasoning. They find that frontier models can exhibit invisible reasoning by leveraging semantically irrelevant filler tokens. In other words, invisible reasoning can ser
Outlook where frontier AI is headed next 18 months: The AI reasoning training + harness loop works if you can produce enough data and reasoning traces (via verifiers). Proven with code and math result
Luna non-reasoning is a bit better than GPT 4o which was sota 2 years ago. Luna (medium) thinking is a bit better than GPT-5 (High) which was sota 1 year ago. Now free to everyone unlimited Sol Max/Fa
OpenAI has released GPT‑5.6 Sol, which powers both instant and deep‑reasoning modes for ChatGPT Plus and Pro customers, delivering fact‑centric responses. Free and Go tier users will receive unlimited
OpenAI announced that Plus and Pro subscribers now have a slider to adjust the amount of reasoning effort ChatGPT applies to each response. The update employs GPT‑5.6 Sol for both Instant and deep rea
We’re making better intelligence easier to access in ChatGPT for everyone: - GPT-5.6 Sol now powers both Instant and deep reasoning for Plus & Pro users, delivering more factual, focused responses. -
OpenAI and Anthropic’s language models keep their internal reasoning encrypted; what users see in the UI is only a filtered summary of that reasoning. This practice was highlighted in a tweet by Sarah
Nice primer on post-training reasoning data. (bookmark it) This is one of the first primers to pull the scattered post-training reasoning-data literature into one place, synthesizing over 150 public s
95% token reduction. 30x faster execution. 90%+ task completion. Today at #MSBuild, we announced a major shift to move reasoning upstream: Pinecone Nexus now integrates directly with @Microsoft OneLak
This post likely discusses an innovative reward function design that optimizes for reasoning efficiency in AI systems, possibly in the context of language models or reinforcement learning. The entry a
arXiv:2602.11675v3 Announce Type: replace Abstract: Large language models may answer causal questions correctly for the wrong reasons, substituting associational shortcuts P(Y|X) for the interventiona
Anthropic says Opus 4.7 hits 80.6% on Document Reasoning — up from 57.1%. But 'reasoning about documents' ≠ 'parsing documents for agents.' We ran it on ParseBench. → Charts: 13.5% → 55.8% (+42.3) — h
Claude Opus 4.7 is now available in Windsurf 2.0! Anthropic has clearly optimized Claude Opus 4.7 for sustained reasoning over long runs. Agents stay on track longer without intervention, so engineers
Cursor has integrated Claude Opus 4.7 into its platform, highlighting the model's impressive autonomous capabilities and enhanced creative reasoning abilities. The announcement suggests a new feature
LLM agents loop, drift, and get stuck on hard reasoning tasks up to 30% of the time. Current fixes are either too blunt (hard step limits) or too expensive (LLM-as-judge adding 10-15% overhead per ste
// When to Retrieve During Reasoning // Pay attention to this one, AI devs. (bookmark it) Most RAG systems retrieve once, before the model starts reasoning. Large reasoning models like o1 and R1 don't
NEW: Meta announces Muse Spark. All you need to know: * It's their new multi-modal reasoning model. * Strong at multi-agent orchestration and multi-modal reasoning. * Contemplating mode orchestrates m
Today, we’re launching Alpamayo 2 Super, our frontier open reasoning model for autonomous vehicles. Beyond seeing, Alpamayo understands and reasons through the complex world - thinks before it acts. I
Reasoning from scratch round 2: In this video, I cover the text generation process in LLMs and KV caching (to prepare the base model before adding reasoning techniques in the upcoming ones). 00:00 Int
'The important thing is not to stop questioning.' — Einstein, 71 years on. Today we celebrate him with II-Agent. His genius was reasoning from first principles, a refusal to take any foundation on fai
Banger paper from Stanford, Together AI, and colleagues. Why is this an interesting paper? Today's LLM agents must write out every reasoning step as text before they can search, run code, or take acti
Very cool to see Embedding Gemma 2 by @Google, number on trending on HF! The media and silicon valley bubble only talks about reasoning LLMs when there is a much bigger variety of AI models that are s
Reasoning From Scratch: Reinforcement Learning with Verifiable Rewards (RLVR) round 2. Covering clipped policy ratios, KL loss term, format rewards, and other GRPO tips & tricks. 00:00 Introduction an
**Wild MISSINGNO.** Nous Research has released “Wild MISSINGNO.”, a coding‑and‑agentic reasoning model that emphasizes high token efficiency. It is available for free on the Nous Portal for a limited
A consequence of how frontier models are trained is motivated reasoning, a phenomenon well studied in humans and discussed in this podcast from @PalisadeAI. Our thoughts, beliefs and reasonings tend t
The way most people leverage test-time compute today is via a form of test-time NL reasoning that is computationally equivalent to test-time search (sometimes with a verifier / grader in the loop). Th
// Programmatic Memory Enables Long-Horizon Reasoning // Keep the entire interaction log and search it. It works great and beats bespoke memory harnesses on long-horizon tasks. New research introduces
Another banger paper from Microsoft. Why it's a big deal: It teaches reasoning models to compress their own chain-of-thought mid-generation. The most interesting finding isn't the 2-3x memory savings
Highlights: 👉 Configurable thinking mode for step-by-step reasoning 👉 Multimodal understanding with text and image input, including document parsing and OCR 👉 Native function calling with structured t
Reasoning from scratch, round number 6! An introduction (and implementation) of Reinforcement Learning with Verifiable Rewards (RLVR) and Group Relative Policy Optimization (GRPO). 00:00 Introduction
I said 'auto-regressive LLMs, in and of themselves, will not lead human-level AI' That statement is still totally true. First, the reasoning abilities of current AI systems are based non-auto-regressi
Great tips on working with reasoning models. Normally you would append what the model figured out after the document and ask again. It turns out that where you put the reasoning trace changes long-con
As OpenAI (and presumably Chinese open source model makers) flirt with 'neuralese,' making so-called 'chain of thought' reasoning non-auditable, the need to refocus efforts on neurosymbolic approaches
Must-read research by Anthropic. Here is the simple explanation and why this is a big deal. We suspect LLMs perform 'internal reasoning'. But little is known or do good methods exist to understand it.
This post discusses how observing Claude's internal activation patterns (J-space) reveals the model's hidden reasoning processes, such as detecting code bugs and analyzing images, even when these step
NEW paper worth reading. Reasoning-data curation is expensive because scoring a trace usually means reading it to the end. This new work from UCLA shows you may not have to. The quality of a reasoning
Reasoning from scratch, round number 5! This time, talking about log-probability scoring (also a great fundamental concept for loss functions like cross-entropy in pre-training and distillation) and s
Reasoning from scratch round 3: This time, I cover generating a verifier for... a) ...evaluation (base model versus any future model improvement) b) ...the reinforcement learning with verifiable rewar
// Agentic Context Management // Great read for the weekend. (bookmark it) Production agents fail less on reasoning and more on what sits in their context. Conversation history, big prompts, huge tool
We stress tested many frontier AI models for multimodal medical reasoning (including GPT-5, Claude 3.5, Gemini 2.5 Pro). They’re not ready. Faulty reasoning, use of inappropriate shortcuts, hallucinat
EBM are so back! @ylecun has been pointing here for years: AI reasoning needs systems that check structure before they answer. Aleph from @logic_int now leads the major formal reasoning benchmarks – l
Energy-Based Models (EBMs) have demonstrated renewed competitiveness in formal reasoning tasks, with Aleph achieving leading performance on major benchmarks in this domain. This represents a significa
Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents. Voice agents are now real-time collaborators that can listen, reason, and s
Today we're releasing ZAYA1-8B, a reasoning MoE trained on @AMD and optimized for intelligence density. With <1B active params, it outperforms open-weight models many times its size on math and reason
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western open frontier on coding & agentic tasks - Trai
Very good to see more competent labs training open-weight models. This Reflection's new Beam model seems to have strong reasoning efficiency, which I think is a big deal for long-running agents. 3-4x
Next year frontier models are going to one shot everything 100x faster & cheaper than they are No reasoning traces to monitor Instinctive intelligence Which is why what you embed in pretraining is eve
Big new release of my LLM CLI tool and Python library for talking to hundreds of different LLMs - reasoning traces, OpenAI Responses support, server-side tools, smarter logging and a whole lot more ht
This video from the World Science Festival, shared by cognitive scientist Gary Marcus, examines the gap between AI systems' claimed reasoning capabilities and their actual mechanisms, likely arguing t
🚀 Meet Qwen3.8-Omni-Flash, Qwen's first omni-modal model built around agentic capabilities! Native audio-video understanding, reasoning, and tool use come together in one model: understand the content
Top open models by use case 👇 chat - Kimi K3, Qwen3.8 2.4T A95B reasoning - Kimi K3, DeepSeek V4 Pro coding agents - Kimi K3, DeepSeek V4 Flash, GLM 5.2 small + fast - Qwen3.8 27B, gpt-oss 20B mid-siz
Our general-purpose coding agent just scored 100% on the ARC-AGI-3 interactive reasoning benchmark. NVIDIA AVO completed all 183 levels across all 25 public environments, figuring out what to do with
8/19, join us for the #ACMTechTalk, 'From Conventional LLMs to Reasoning Models to Agents,' w/AI & LLM Research Engineer @rasbt. ACM Practitioner Board Co-Chaior @marlene_zw (@Microsoft) will moderate
One of the most confusing aspects of GPT-5.6 is figuring out which model to use at which reasoning effort - sounds like Sol on Medium might be a good new default for coding work, if it's an upgrade fr
// The Consistency Illusion // Multi-agent debate can make agents agree on the final answer while their underlying reasoning stays misaligned. This work finds that consensus on the output hides disagr
what constitutes reasoning in AI is a critical debate. i hope that @dwarkesh_sp will respond. what does this even mean, @dwarkesh_sp, “the real deal”? is it even a falsifiable conjecture? what’s the e
Mistral Large 4 against the field: efficiently reasoning over diverse complex challenges. CTF speedrun: 19 challenges, with tool calls and solve times grounded in actual runs. It solved 18 of the 19.