mlsanityData to 7 Oct 2026

Fluid Reasoning Representations

arXiv 2602.04843 ↗First posted 4 Feb 2026cs.AIcited by 0

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Frontier large language models increasingly solve complex tasks involving abstract concepts through extended test-time thinking. Yet we lack a mechanistic account of how extended thinking changes hidden-state representations over the course of a reasoning trace. We introduce textitFluid Reasoning Representations (FRRs), a representation-level account of how LLMs organize action and predicate concepts during self-generated reasoning, and test them on obfuscated planning, symbolic, and mathematical tasks where task-relevant words are replaced while problem structure is preserved. Across open-weight base, instruction-tuned, and extended-thinking LLMs, representations of the same action or predicate become more similar across wordings and move toward the corresponding unobfuscated concepts over the reasoning trace. Causal probes show that these representations affect behavior: cross-naming steering improves held-out accuracy beyond Gaussian and shuffled controls, symbolic patching retains more action information than shuffled patching, and subtracting refined directions degrades accuracy; together, these results suggest that extended thinking strengthens a representation dynamic already present at lower magnitude in base and instruction-tuned LLMs. Our codebase is open-sourced hrefhttps://github.com/AI4Collaboration/Fluid-Reasoning-Representationhere.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 11 of its 20 references that are in the map. First posted on arXiv 4 Feb 2026.

Builds on

11 of 20 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models16 Jan

Other references · 10

  1. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  2. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  3. 2022PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Changecited by 372
  4. 2025Thought Anchors: Which LLM Reasoning Steps Matter?cited by 126
  5. 2024LLMs Still Can't Plan; Can LRMs? A Preliminary Evaluation of OpenAI's o1 on PlanBenchcited by 93
  6. 2025Understanding Reasoning in Thinking Language Models via Steering Vectorscited by 69
  7. 2024ICLR: In-Context Learning of Representationscited by 49
  8. 2024NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internalscited by 43
  9. 2025Reasoning-Finetuning Repurposes Latent Representations in Base Modelscited by 23
  10. 2025The Geometry of Self-Verification in a Task-Specific Reasoning Modelcited by 12

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Reasoning as State Transition: A Representational Analysis of Reasoning Evolution in Large Language Models
  2. 2025Cognitive Foundations for Reasoning and Their Manifestation in LLMs
  3. 2025Truly Assessing Fluid Intelligence of Large Language Models through Dynamic Reasoning Evaluation
  4. 2025Mapping the Minds of LLMs: A Graph-Based Analysis of Reasoning LLM
  5. 2025Beyond Pattern Recognition: Probing Mental Representations of LMs
  6. 2026PRISM: A Dual View of LLM Reasoning through Semantic Flow and Latent Computation
  7. 2025Scaling Competence, Shrinking Reasoning: Cognitive Signatures in Language Model Learning
  8. 2025Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning