mlsanityData to 7 Oct 2026

Characterizing, Evaluating, and Optimizing Complex Reasoning

arXiv 2602.08498 ↗First posted 9 Feb 2026cs.CLcited by 5

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

Abstract (arXiv)

Large Reasoning Models (LRMs) increasingly rely on reasoning traces with complex internal structures. However, existing work lacks a unified answer to three fundamental questions: (1) what defines high-quality reasoning, (2) how to reliably evaluate long, implicitly structured reasoning traces, and (3) how to use such evaluation signals for reasoning optimization. To address these challenges, we provide a unified perspective. (1) We introduce the ME2 principle to characterize reasoning quality along macro- and micro-level concerning efficiency and effectiveness. (2) Built on this principle, we model reasoning traces as directed acyclic graphs (DAGs) and develop a DAG-based pairwise evaluation method, capturing complex reasoning structures. (3) Based on this method, we construct the TRM-Preference dataset and train a Thinking Reward Model (TRM) to evaluate reasoning quality at scale. Experiments show that thinking rewards serve as an effective optimization signal. At test time, selecting better reasoning leads to better outcomes (up to 19.3% gain), and during RL training, thinking rewards enhance reasoning and performance (up to 3.9% gain) across diverse tasks. Code and data are available at https://github.com/Simplified-Reasoning/TRM.

Key facts. Cited by 5 papers 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 40 of its 65 references that are in the map. First posted on arXiv 9 Feb 2026.

Builds on

40 of 65 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025Reasoning Efficiently Through Adaptive Chain-of-Thought Compression: A Self-Optimizing Framework17 Sep
  2. 2025gpt-oss-120b & gpt-oss-20b Model Card8 Aug
  3. 2025Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropy5 Aug
  4. 2025Reconsidering Overthinking: Penalizing Internal and External Redundancy in CoT Reasoning4 Aug
  5. 2025ReasoningFlow: Semantic Structure of Complex Reasoning Traces3 Jun
  6. 2025Can Pruning Improve Reasoning? Revisiting Long-CoT Compression with Capability in Mind for Better Reasoning20 May

Other references · 34

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  4. 2025Qwen3 Technical Reportcited by 8,195
  5. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  6. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
  7. 2024OpenAI o1 System Cardcited by 1,220
  8. 2024Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvementcited by 1,057
  9. 2021A General Language Assistant as a Laboratory for Alignmentcited by 1,023
  10. 2025DeepSeek-V3.2: Pushing the Frontier of Open Large Language Modelscited by 755
  11. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 505
  12. 2025The Lessons of Developing Process Reward Models in Mathematical Reasoningcited by 402
22 more
  1. 2023Understanding the Effects of RLHF on LLM Generalisation and Diversitycited by 386
  2. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 290
  3. 2025Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMscited by 276
  4. 2025Learning to Reason under Off-Policy Guidancecited by 266
  5. 2024Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judgecited by 249
  6. 2025Why Language Models Hallucinatecited by 205
  7. 2025General-Reasoner: Advancing LLM Reasoning Across All Domainscited by 132
  8. 2024Is LLM-as-a-Judge Robust? Investigating Universal Adversarial Attacks on Zero-shot LLM Assessmentcited by 117
  9. 2025DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoningcited by 84
  10. 2025SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplinescited by 67
  11. 2024A Chain-of-Thought Is as Strong as Its Weakest Link: A Benchmark for Verifiers of Reasoning Chainscited by 63
  12. 2025Evaluating Step-by-step Reasoning Traces: A Surveycited by 62
  13. 2025ExGRPO: Learning to Reason from Experiencecited by 49
  14. 2025ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMscited by 44
  15. 2025R-PRM: Reasoning-Driven Process Reward Modelingcited by 38
  16. 2025Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Propertiescited by 28
  17. 2025Mapping the Minds of LLMs: A Graph-Based Analysis of Reasoning LLMcited by 21
  18. 2024General Purpose Verification for Chain of Thought Promptingcited by 16
  19. 2024Structured Chain-of-Thought Prompting for Few-Shot Generation of Content-Grounded QA Conversationscited by 9
  20. 2025ReasonGraph: Visualisation of Reasoning Pathscited by 6
  21. 2025Confidence and Stability of Global and Pairwise Scores in NLP Evaluationcited by 4
  22. 2025DAG-Math: Graph-of-Thought Guided Mathematical Reasoning in LLMscited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Tail-Aware Top-k On-Policy Distillationcited by 3
  2. 2026TRACE: A Self-Evolving Skill Bank for Consistent, Limit-Aware LLM Agentscited by 2
  3. 2026ComBench: A Benchmark for Rigorous Proof Reasoning and Constructive Realization in Olympiad-Level Combinatoricscited by 0
  4. 2026LCoT-GV: Graph Attention Networks for Verifying Long Reasoning Chains in Large Language Modelscited by 0
  5. 2026RGDT-Bench: Benchmarking LLM Reasoning for Rule-Governed Decisions and Their Justificationscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
  2. 2025THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models
  3. 2026InfoDensity: Rewarding Information-Dense Traces for Efficient Reasoning
  4. 2025Rethinking Reasoning Quality in Large Language Models through Enhanced Chain-of-Thought via RL
  5. 2025Evaluating the Logical Reasoning Abilities of Large Reasoning Models
  6. 2025Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoning
  7. 2025Logical Reasoning with Outcome Reward Models for Test-Time Scaling
  8. 2026Dynamic Thinking-Token Selection for Efficient Reasoning in Large Reasoning Models