mlsanityData to 14 Oct 2026

Intention Collapse: Intention-Level Metrics for Reasoning in Language Models

arXiv 2601.01011 ↗First posted 3 Jan 2026cs.CL cs.AIcited by 0

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Language generation maps a rich, high-dimensional internal state to a single token sequence. We study this many-to-one mapping through the lens of intention collapse: the projection from an internal intention space I to an external language space L. We introduce three cheap, model-agnostic metrics computed on a pre-collapse state I: (i) intention entropy Hint(I), (ii) effective dimensionality deff(I), and (iii) recoverability Recov(I), operationalized as probe AUROC for predicting eventual success. We evaluate these metrics in a 3x3 study across models (Mistral-7B, LLaMA-3.1-8B, Qwen-2.5-7B) and benchmarks (GSM8K, ARC-Challenge, AQUA-RAT), comparing baseline, chain-of-thought (CoT), and a babble control (n=200 items per cell). CoT increases average accuracy from 34.2% to 47.3% (+13.1 pp), driven by large gains on GSM8K but consistent degradations on ARC-Challenge. Across models, CoT induces distinct entropy regimes relative to baseline, dH = Hint(CoT) - Hint(Base): Mistral shows dH < 0 (lower-entropy CoT), whereas LLaMA shows dH > 0 (higher-entropy CoT), highlighting heterogeneity in CoT-induced internal uncertainty. Finally, probe AUROC is significantly above chance in a subset of settings and can dissociate from behavioral accuracy (e.g., high AUROC alongside lower CoT accuracy on ARC-Challenge for Qwen), suggesting that informative internal signal is not always reliably converted into a final discrete decision under constrained response formats.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 27 of its 40 references that are in the map. First posted on arXiv 3 Jan 2026.

Builds on

27 of 40 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters6 Aug
  2. 2023Self-Refine: Iterative Refinement with Self-Feedback30 Mar
  3. 2022Language Models (Mostly) Know What They Know11 Jul

Other references · 24

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  5. 2023Let's Verify Step by Stepcited by 3,789
  6. 2024Qwen2 Technical Reportcited by 2,120
  7. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,122
  8. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 991
  9. 2022Solving math word problems with process- and outcome-based feedbackcited by 901
  10. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 819
  11. 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedbackcited by 812
  12. 2022Discovering Latent Knowledge in Language Models Without Supervisioncited by 767
12 more
  1. 2022Teaching Models to Express Their Uncertainty in Wordscited by 751
  2. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 724
  3. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 703
  4. 2024Quiet-STaR: Language Models Can Teach Themselves to Think Before Speakingcited by 301
  5. 2024Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoningcited by 279
  6. 2022Test-Time Training with Masked Autoencoderscited by 212
  7. 2023Let's reward step by step: Step-Level reward model as the Navigators for Reasoningcited by 102
  8. 2024The Surprising Effectiveness of Test-Time Training for Few-Shot Learningcited by 96
  9. 2023SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inferencecited by 89
  10. 2023Eliciting Latent Knowledge from Quirky Language Modelscited by 62
  11. 2022Can language models handle recursively nested grammatical structures? A case study on comparing models and humanscited by 41
  12. 2024Improving Complex Reasoning over Knowledge Graph with Logic-Aware Curriculum Tuningcited by 10

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Dissecting Failure Dynamics in Large Language Model Reasoning
  2. 2026Reasoning as State Transition: A Representational Analysis of Reasoning Evolution in Large Language Models
  3. 2025State over Tokens: Characterizing the Role of Reasoning Tokens
  4. 2025SWI: Speaking with Intent in Large Language Models
  5. 2026Large Language Models Decide Early and Explain Later
  6. 2025Looking beyond the next token
  7. 2025On the Limits of Innate Planning in Large Language Models
  8. 2025Temporal Predictors of Outcome in Reasoning Language Models