Preprint · 2025 · Open access

Revisiting LLM Reasoning via Information Bottleneck

Shiye Lei, Zhihao Cheng, Jia, Kai, Dacheng Tao

arXiv (Cornell University) · Published 24 July 2025 · DOI 10.48550/arxiv.2507.18391

Free PDFView at the publisherCite this paper

In brief

“Revisiting LLM Reasoning via Information Bottleneck” is a 2025 preprint by Shiye Lei and 3 co-authors, published in arXiv (Cornell University). OpenAlex counts 22 citations as of 15 November 2026. It is open access: the full text is free to read at arXiv (Cornell University).

Citations
22
Most cited in
2026 (14)
References
18
Authors
4

Abstract

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.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Lei, S., Cheng, Z., Kai, J., & Tao, D. (2025). Revisiting LLM Reasoning via Information Bottleneck. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2507.18391

MLA 9

Lei, Shiye, et al. “Revisiting LLM Reasoning via Information Bottleneck.” arXiv (Cornell University), 2025. https://doi.org/10.48550/arxiv.2507.18391.

Chicago

Lei, Shiye, Zhihao Cheng, Jia, Kai, and Dacheng Tao. 2025. “Revisiting LLM Reasoning via Information Bottleneck.” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2507.18391.

Harvard

Lei, S. et al. (2025) ‘Revisiting LLM Reasoning via Information Bottleneck’, arXiv (Cornell University). doi:10.48550/arxiv.2507.18391.

IEEE

S. Lei, Z. Cheng, J. Kai, and D. Tao, “Revisiting LLM Reasoning via Information Bottleneck,” arXiv (Cornell University), 2025, doi: 10.48550/arxiv.2507.18391.

BibTeX

@misc{lei2025revisiting,
  title     = {{Revisiting LLM Reasoning via Information Bottleneck}},
  author    = {Lei, Shiye and Cheng, Zhihao and Kai, Jia, and Tao, Dacheng},
  howpublished = {arXiv (Cornell University)},
  year      = {2025},
  doi       = {10.48550/arxiv.2507.18391},
  url       = {https://doi.org/10.48550/arxiv.2507.18391}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling

  • large language models
  • chain-of-thought reasoning
  • information bottleneck
  • reinforcement learning with verifiable rewards
  • mathematical reasoning
  • post-training fine-tuning

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Lei, S., Cheng, Z., Kai, J., & Tao, D. (2025). Revisiting LLM Reasoning via Information Bottleneck. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2507.18391

How many times has this paper been cited?

22 times, according to OpenAlex on 15 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access: read the free full text at arXiv (Cornell University).

Who wrote this paper?

Shiye Lei, Zhihao Cheng, Jia, Kai, Dacheng Tao.

What is the DOI of this paper?

10.48550/arxiv.2507.18391. The permanent link is https://doi.org/10.48550/arxiv.2507.18391.

Bibliographic data from OpenAlex (CC0), last checked 15 November 2026. ResearchGuru does not host this paper; the links go to the publisher or the repository that does.