Pith. sign in

REVIEW 12 cited by

Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2506.02867 v2 pith:4R66TFTA submitted 2025-06-03 cs.AI cs.CL

Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning

classification cs.AI cs.CL
keywords reasoningtokensthinkinginformationlrmspeakscapabilitiesduring
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Large reasoning models (LRMs) have demonstrated impressive capabilities in complex problem-solving, yet their internal reasoning mechanisms remain poorly understood. In this paper, we investigate the reasoning trajectories of LRMs from an information-theoretic perspective. By tracking how mutual information (MI) between intermediate representations and the correct answer evolves during LRM reasoning, we observe an interesting MI peaks phenomenon: the MI at specific generative steps exhibits a sudden and significant increase during LRM's reasoning process. We theoretically analyze such phenomenon and show that as MI increases, the probability of model's prediction error decreases. Furthermore, these MI peaks often correspond to tokens expressing reflection or transition, such as ``Hmm'', ``Wait'' and ``Therefore,'' which we term as the thinking tokens. We then demonstrate that these thinking tokens are crucial for LRM's reasoning performance, while other tokens has minimal impacts. Building on these analyses, we propose two simple yet effective methods to improve LRM's reasoning performance, by delicately leveraging these thinking tokens. Overall, our work provides novel insights into the reasoning mechanisms of LRMs and offers practical ways to improve their reasoning capabilities. The code is available at https://github.com/ChnQ/MI-Peaks.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Beyond Entropy: Learning from Token-Level Distributional Deviations for LLM Reasoning

    cs.AI 2026-06 unverdicted novelty 7.0

    ICT framework applies JS divergence to token logits to select critical tokens for selective RLVR updates, claiming 4.58% average pass@4 gains on Qwen2.5 models across seven reasoning benchmarks.

  2. How to Fine-Tune a Reasoning Model? A Teacher-Student Cooperation Framework to Synthesize Student-Consistent SFT Data

    cs.CL 2026-03 conditional novelty 7.0

    TESSY creates stylistically consistent synthetic data via teacher-student token interleaving, yielding 11.25% and 6.68% gains on code benchmarks where pure teacher data causes 3.25% and 10.02% drops.

  3. The Stepwise Informativeness Assumption: Why are Entropy Dynamics and Reasoning Correlated in LLMs?

    cs.CL 2026-03 unverdicted novelty 7.0

    The Stepwise Informativeness Assumption explains the correlation between LLM entropy dynamics and reasoning correctness by positing that correct traces accumulate answer-relevant information stepwise during generation.

  4. Large Reasoning Models Are (Not Yet) Multilingual Latent Reasoners

    cs.CL 2026-01 unverdicted novelty 7.0

    Large reasoning models exhibit multilingual latent reasoning that is uneven across languages but internally consistent and English-centered.

  5. Attention Degradation, Function Token Anchoring, and the Limits of Attention-Based Intervention in Large Language Models

    cs.AI 2026-07 conditional novelty 6.0

    In GPT-2, OPT-1.3B, and LLaMA-3.2 models, boosting attention to function-token positions by 16-24% does not improve contextual retrieval, so short-range attention degradation is descriptive rather than causal.

  6. The Tell-Tale Norm: $\ell_2$ Magnitude as a Signal for Reasoning Dynamics in Large Language Models

    cs.CL 2026-06 unverdicted novelty 6.0

    The L2 norm of LLM hidden states signals reasoning intensity, with a theoretical bound on SAE feature activations, enabling three new test-time scaling techniques that boost performance.

  7. Attention-Guided Reward for Reinforcement Learning-based Jailbreak against Large Reasoning Models

    cs.AI 2026-05 unverdicted novelty 6.0

    An attention-guided RL reward combined with diverse persuasion strategies produces higher attack success rates against large reasoning models than prior jailbreak methods.

  8. Preference Tuning as Spectral Update Reorganization

    cs.CL 2026-05 conditional novelty 6.0

    Preference tuning reorganizes a model's weight update into a dominant spectral head that drives visible behavior and a residual tail that supports learning breadth; the update is structured, not monolithic.

  9. Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding

    cs.AI 2026-05 unverdicted novelty 6.0

    CoRD uses collaborative multi-teacher step-wise decoding with perplexity-guided beam search to generate higher-quality Long-CoT data that lets smaller models reach near-teacher performance with less supervision.

  10. Thinking Seeds: Leveraging Historical Diversity for Position-Aware RL in LLMs

    cs.CL 2026-01 conditional novelty 6.0

    SOUP mixes off-policy historical prefixes with on-policy continuations at token level and reports small but consistent math-reasoning gains over on-policy GRPO/DAPO baselines in selected configurations.

  11. "The Whole Is Greater Than the Sum of Its Parts": A Compatibility-Aware Multi-Teacher CoT Distillation Framework

    cs.CL 2026-01 unverdicted novelty 6.0

    COMPACT adaptively fuses multi-teacher CoT supervisions using graph-based consensus, mutual-information adaptability, and loss-based difficulty metrics to improve small language model reasoning performance while mitig...

  12. Early Stopping Chain-of-thoughts in Large Language Models

    cs.CL 2025-09 conditional novelty 5.0

    ES-CoT shortens LLM chain-of-thought generation by tracking runs of identical step answers after linguistic markers, cutting tokens 16% on average while keeping accuracy comparable to full CoT across six datasets and ...