Is my model "mind blurting"? Interpreting the dynamics of reasoning tokens with Recurrence Quantification Analysis (RQA)
On the map: LLM reasoning → Test-time compute and efficient thinking
Abstract (arXiv)
Test-time compute is central to large reasoning models, yet analysing their reasoning behaviour through generated text is increasingly impractical and unreliable. Response length is often used as a brute proxy for reasoning effort, but this metric fails to capture the dynamics and effectiveness of the Chain of Thoughts (CoT) or the generated tokens. We propose Recurrence Quantification Analysis (RQA) as a non-textual alternative for analysing model's reasoning chains at test time. By treating token generation as a dynamical system, we extract hidden embeddings at each generation step and apply RQA to the resulting trajectories. RQA metrics, including Determinism and Laminarity, quantify patterns of repetition and stalling in the model's latent representations. Analysing 3,600 generation traces from DeepSeek-R1-Distill, we show that RQA captures signals not reflected by response length, but also substantially improves prediction of task complexity by 8%. These results help establish RQA as a principled tool for studying the latent token generation dynamics of test-time scaling in reasoning models.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 13 of its 23 references that are in the map. First posted on arXiv 5 Feb 2026.
Builds on
13 of 23 references.
Closest ancestors · 3
Same direction, newest first
- 2025Understanding Reasoning in Thinking Language Models via Steering Vectors22 Jun
- 2025A Statistical Physics of Language Model Reasoning4 Jun
- 2025ThinkEdit: Interpretable Weight Editing to Mitigate Overly Short Thinking in Reasoning Models27 Mar
Other references · 10
- 2025ZebraLogic: On the Scaling Limits of LLMs for Logical Reasoningcited by 186
- 2025Thought Anchors: Which LLM Reasoning Steps Matter?cited by 126
- 2024How to think step-by-step: A mechanistic understanding of chain-of-thought reasoningcited by 53
- 2023What's the Magic Word? A Control Theory of LLM Promptingcited by 41
- 2025Position: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!cited by 36
- 2024Diffusion of Thoughts: Chain-of-Thought Reasoning in Diffusion Language Modelscited by 26
- 2025How does Chain of Thought Think? Mechanistic Interpretability of Chain-of-Thought Reasoning with Sparse Autoencodingcited by 19
- 2024Autoregressive + Chain of Thought = Recurrent: Recurrence's Role in Language Models' Computability and a Revisit of Recurrent Transformercited by 16
- 2025Exploring Information Processing in Large Language Models: Insights from Information Bottleneck Theorycited by 8
- 2025Translate Policy to Language: Flow Matching Generated Rewards for LLM Explanationscited by 6
Built on by
Citing papers per year: 2026: 1.
1 paper.
- 2026Interpreting Latent CoT Reasoning as Dynamical Systemscited by 0
- 2026Interpreting Latent CoT Reasoning as Dynamical Systems20 Jun 2026
Similar, not cited
Close in wording, no citation either way
- 2025The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis
- 2026Latent Recurrent Thoughts: Recurrent Refinement of Proposed Latents for Reasoning with Frozen LLMs
- 2026A Mechanistic Analysis of Looped Reasoning Language Models
- 2025Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoning
- 2025Thought calibration: Efficient and confident test-time scaling
- 2025Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Time
- 2026Answer-Distribution Trajectories: A Stochastic-Dynamics View of LLM Reasoning
- 2025Scaling Competence, Shrinking Reasoning: Cognitive Signatures in Language Model Learning