mlsanityData to 7 Oct 2026

LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition

arXiv 2605.24005 ↗First posted 19 May 2026cs.AI cs.CLcited by 0

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

also in 1 more
Abstract (arXiv)

The evolution of Large Language Model (LLM) reasoning is bottlenecked by the scarcity of high-quality process data. While self-alignment via endogenous rewards offers a solution, mining valid supervision faces three challenges: (1) Label Noise via Mimetic Bias, where rewards prioritize statistical likelihood over logical truth, creating a "correctness illusion" that masks compounding errors; (2) Coarse-Grained Supervision, where sparse global outcomes (e.g., in GRPO) fail to provide granular guidance, treating reasoning chains as monolithic; and (3) Distributional Collapse, where signals fail to generalize without amplifying pre-training biases. To address these, we introduce LC-ERD (Logic-Consistent Endogenous Reward Decomposition), a framework framing self-alignment as latent structure mining. We derive a Variational Logic Potential by aggregating consensus from the model's Latent Logic Expertise (LLE) to denoise the reasoning manifold, and introduce a Multi-Agent Value Decomposition protocol based on the IGM principle to quantify individual step utility. Experiments show LC-ERD delivers a robust self-evolution path, uncovering trade-offs between logic consistency and accuracy while identifying high-value reasoning patterns missed by standard rewards. Our code is available at https://github.com/LC-ERD-repo/LC-ERD.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 26 of its 50 references that are in the map. First posted on arXiv 19 May 2026.

Builds on

26 of 50 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs5 Feb

Other references · 25

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2023Safe RLHF: Safe Reinforcement Learning from Human Feedbackcited by 630
  4. 2024AFlow: Automating Agentic Workflow Generationcited by 395
  5. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  6. 2024Is DPO Superior to PPO for LLM Alignment? A Comprehensive Studycited by 262
  7. 2021Maximum Entropy RL (Provably) Solves Some Robust RL Problemscited by 169
  8. 2025InfiGUI-R1: Advancing Multimodal GUI Agents from Reactive Actors to Deliberative Reasonerscited by 140
  9. 2024DPO Meets PPO: Reinforced Token Optimization for RLHFcited by 126
  10. 2023A Survey of Temporal Credit Assignment in Deep Reinforcement Learningcited by 56
  11. 2025ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMscited by 44
  12. 2025Generalist Reward Models: Found Inside Large Language Modelscited by 34
13 more
  1. 2019Learning to solve the credit assignment problemcited by 27
  2. 2025Optimizing Safe and Aligned Language Generation: A Multi-Objective GRPO Approachcited by 24
  3. 2025MO-GRPO: Mitigating Reward Hacking of Group Relative Policy Optimization on Multi-Objective Problemscited by 21
  4. 2024MedQA-CS: Objective Structured Clinical Examination (OSCE)-Style Benchmark for Evaluating LLM Clinical Skillscited by 16
  5. 2025On Group Relative Policy Optimization Collapse in Agent Search: The Lazy Likelihood-Displacementcited by 15
  6. 2021Offline Inverse Reinforcement Learningcited by 14
  7. 2025TIC-GRPO: Provable and Efficient Optimization for Reinforcement Learning from Human Feedbackcited by 11
  8. 2025Quick on the Uptake: Eliciting Implicit Intents from Human Demonstrations for Personalized Mobile-Use Agentscited by 11
  9. 2025MATH-Beyond: A Benchmark for RL to Expand Beyond the Base Modelcited by 10
  10. 2025Why Tree-Style Branching Matters for Thought Advantage Estimation in GRPOcited by 9
  11. 2026Agent-Dice: Disentangling Knowledge Updates via Geometric Consensus for Agent Continual Learningcited by 6
  12. 2025Diffusion Fine-Tuning via Reparameterized Policy Gradient of the Soft Q-Functioncited by 5
  13. 2026Sub-optimality bounds for certainty equivalent policies in partially observed systemscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding
  2. 2026Miner:Mining Intrinsic Mastery for Data-Efficient RL in Large Reasoning Models
  3. 2025Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models
  4. 2025A Survey on Latent Reasoning
  5. 2026Latent Thought Flow: Efficient Latent Reasoning in Large Language Models
  6. 2025Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning
  7. 2025SpiralThinker: Latent Reasoning through an Iterative Process with Text-Latent Interleaving
  8. 2026Emergent Search and Backtracking in Latent Reasoning Models