Reasoning and the "physics" of language models — page 1
- A Formal Comparison Between Chain of Thought and Latent Thought
arxiv:2509.25239· 2025-09-25 · text available - A Mechanistic Analysis of Looped Reasoning Language Models
arxiv:2604.11791· 2026-04-13 · text available - AI-rithmetic
arxiv:2602.10416· 2026-02-11 · text available - Are Emergent Abilities of Large Language Models a Mirage?
arxiv:2304.15004· 2023-04-28 · text available - Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics
arxiv:2410.21272· 2024-10-28 · text available - Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers
arxiv:2606.31779· 2026-06-30 · text available - Can Large Reasoning Models Self-Train?
arxiv:2505.21444· 2025-05-27 · text available - Characterizing emergent phenomena in LLMs (Google Research)
url:ai.googleblog.com/2022/11/characterizing-emergent-phenomena-in.html· 2022-11-10 · metadata only - Competitive Programming with Large Reasoning Models
arxiv:2502.06807· 2025-02-03 · text available - Dissociating language and thought in large language models
arxiv:2301.06627· 2023-01-16 · text available - Emergent Analogical Reasoning in Large Language Models
arxiv:2212.09196· 2022-12-19 · text available - Emergent Capabilities Arise Randomly from Learning Sparse Attention Patterns
arxiv:2606.25010· 2026-06-23 · text available - Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning
arxiv:2605.21488· 2026-05-20 · text available - Evaluating the World Model Implicit in a Generative Model
arxiv:2406.03689· 2024-06-06 · text available - Evidence from formal logical reasoning reveals that the language of thought is not natural language
doi:10.1073/pnas.2520095123· 2026-07-06 · text available - Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformers
arxiv:2606.18206· 2026-06-16 · text available - Flow Reasoning Models: Turning Flows Into Efficient Recurrent Reasoners
arxiv:2606.29150· 2026-06-28 · text available - From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step
arxiv:2405.14838· 2024-05-23 · text available - Generative Recursive Reasoning
arxiv:2605.19376· 2026-05-19 · text available - Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization
arxiv:2405.15071· 2024-05-23 · text available - Grokking Group Multiplication with Cosets
openreview:hcQfTsVnBo· 2023-12-11 · text available - Hierarchical Reasoning Model
arxiv:2506.21734· 2025-06-26 · text available - How do language models learn facts? Dynamics, curricula and hallucinations
arxiv:2503.21676· 2025-03-27 · text available - How Do Large Language Models Acquire Factual Knowledge During Pretraining?
arxiv:2406.11813· 2024-06-17 · text available