mlsanityData to 14 Oct 2026

Revisiting LLM Reasoning via Information Bottleneck

arXiv 2507.18391 ↗First posted 24 Jul 2025cs.AIcited by 19

On the map: LLM reasoning → Reinforcement learning for reasoning

Abstract (arXiv)

Large language models (LLMs) have recently demonstrated remarkable progress in reasoning capabilities through reinforcement learning with verifiable rewards (RLVR). By leveraging simple rule-based rewards, RL effectively incentivizes LLMs to produce extended chain-of-thought (CoT) reasoning trajectories, progressively guiding them toward correct answers. However, existing approaches remain largely heuristic and intuition-driven, limiting the development of principled methodologies. In this paper, we present a theoretical characterization of LLM reasoning grounded in information bottleneck (IB) principle, introducing IB-aware reasoning optimization (IBRO), a framework that encourages reasoning trajectories to be both informative about the final correct answer and generalizable across diverse prompts. We derive a practical token-level surrogate objective and propose an efficient approximation, resulting in the lightweight IB regularization method. This technique integrates seamlessly into existing RL-based post-training frameworks without additional computational overhead, requiring only a one-line code modification. Empirically, we validate IB regularization across multiple mathematical reasoning benchmarks and RL algorithms, demonstrating consistent improvements in LLM reasoning performance.

Key facts. Cited by 19 papers in the map (data through 14 Oct 2026). Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning. Builds on 17 of its 26 references that are in the map. First posted on arXiv 24 Jul 2025.

Builds on

17 of 26 references.

Closest ancestors · 10

Same direction, newest first

  1. 2025Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training16 Jul
  2. 2025Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them13 Jul
  3. 2025Reasoning with Exploration: An Entropy Perspective17 Jun
  4. 2025Confidence Is All You Need: Few-Shot RL Fine-Tuning of Language Models5 Jun
  5. 2025Diversity-Aware Policy Optimization for Large Language Model Reasoning29 May
  6. 2025The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models28 May
  7. 2025The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning21 May
  8. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scale18 Mar
  9. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning22 Jan
  10. 2017Proximal Policy Optimization Algorithms20 Jul

Other references · 7

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2024Qwen2.5 Technical Reportcited by 4,942
  3. 2024OpenAI o1 System Cardcited by 1,244
  4. 2025Skywork Open Reasoner 1 Technical Reportcited by 165
  5. 2025Reinforcement Learning Finetunes Small Subnetworks in Large Language Modelscited by 49
  6. 2025One-shot Entropy Minimizationcited by 32
  7. 2025Memorization-Compression Cycles Improve Generalizationcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 9; 2026: 10.

19 papers.

Sort:

By area:

  1. 2025Breaking the Exploration Bottleneck: Rubric-Scaffolded Reinforcement Learning for General LLM Reasoningcited by 63
  2. 2025No Prompt Left Behind: Exploiting Zero-Variance Prompts in LLM Reinforcement Learning via Entropy-Guided Advantage Shapingcited by 50
  3. 2026Analyzing and Improving Chain-of-Thought Monitorability Through Information Theorycited by 10
  4. 2025Self-Aligned Reward: Towards Effective and Efficient Reasonerscited by 7
  5. 2026A Step Back: Prefix Importance Ratio Stabilizes Policy Optimizationcited by 7
  6. 2025PEAR: Phase Entropy Aware Reward for Efficient Reasoningcited by 5
  7. 2025ReLaX: Reasoning with Latent Exploration for Large Reasoning Modelscited by 5
  8. 2025The Effectiveness of Approximate Regularized Replay for Efficient Supervised Fine-Tuning of Large Language Modelscited by 2
  9. 2026Mitigating Translationese Bias in Multilingual LLM-as-a-Judge via Disentangled Information Bottleneckcited by 2
  10. 2025Topic Identification in LLM Input-Output Pairs through the Lens of Information Bottleneckcited by 1
  11. 2025Breaking Memorization Barriers in LLM Code Fine-Tuning via Information Bottleneck for Improved Generalizationcited by 1
  12. 2026Reasoning as Compression: Unifying Budget Forcing via the Conditional Information Bottleneckcited by 1
7 more
  1. 2026Beyond Entropy: Correctness-Aware Advantage Shaping via Contrastive Policy Optimizationcited by 1
  2. 2025From generative AI to the brain: five takeawayscited by 0
  3. 2026Supervised Fine-Tuning Needs to Unlock the Potential of Token Prioritycited by 0
  4. 2026When Less is Enough: Efficient Inference via Collaborative Reasoningcited by 0
  5. 2026Long Live The Balance: Information Bottleneck Driven Tree-based Policy Optimizationcited by 0
  6. 2026Think Through a Bottleneck: Hourglass Reasoning for Rigorous Inductioncited by 0
  7. 2026Limits of Reliability and Scaling in Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Rectifying LLM Thought from Lens of Optimization
  2. 2026The Art of Efficient Reasoning: Data, Reward, and Optimization
  3. 2026Likelihood-Based Reward Designs for General LLM Reasoning
  4. 2025Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning
  5. 2026The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes
  6. 2025Demystifying Long Chain-of-Thought Reasoning in LLMs
  7. 2025Enhancing Large Language Model Reasoning with Reward Models: An Analytical Survey
  8. 2025AdaR: A Framework for Equipping LLMs with Adaptive Reasoning