Chain of thought

Vol. I No. 26 Saturday, October 17, 2026 Closing 12:53 PM UTC · next edition 10:00 AM UTC

A model writing out its intermediate steps before the answer, which tends to improve results on multi-step problems.

In the news

arXiv cs.CLOct 14, 2026Research
MolLean: A Process-Verifiable Language for Chemical Reasoning
arXiv cs.AIOct 8, 2026Research
Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought Faithfulness
arXiv cs.AIOct 7, 2026Research
Reasoning-Token Spikes Under Prompted Untruthful Responding in Large Language Models
arXiv (vision, retrieval & HCI) cs.CVOct 7, 2026Research
Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving
arXiv cs.AIOct 2, 2026Research
Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability
arXiv cs.CLSep 29, 2026Research
Correct Answers, Invalid Traces: What Verifiable Grade-School Math Reveals About Chain-of-Thought Traces
arXiv cs.AISep 29, 2026Research
Probability is Not Enough: Exploring and Counting Divergent Tokens for Reasoning Uncertainty Quantification in LLMs
arXiv cs.LGSep 29, 2026Research
$S^3$: Spectral Null-Space Swap Makes Reasoning Models Efficient
Apple Machine LearningSep 29, 2026News
The Communication Bottleneck: A Round-Trip Study of Tree-Structured Expression Serialization in Language Models
arXiv cs.CLSep 27, 2026Research
On the Token Value Inequality in Efficient Reasoning
arXiv cs.AISep 25, 2026Research
Monitor Jailbreaking: Evading Chain-of-Thought Monitoring Without Encoded Reasoning
arXiv cs.AISep 24, 2026Research
When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression
arXiv cs.CLSep 24, 2026Research
Reasoning Instructions Can Break Answer Decoding in Vision--Language Models
NVIDIA DeveloperSep 23, 2026News
Introducing NV-Reason-CT Open 3D CT VLM for Radiologist Chain-of-Thought Reasoning
swyx (Latent Space)Sep 23, 2026Columns
Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics)