Conference paper · 2026 · Open access

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

Yanyu Chen, Jiyue Jiang, Dianzhi Yu, Zheng Wu, Jiahong Liu, Jiaming Han, Xiao Guo, Jinhu Qi, Yu Li, Yifei Zhang, Irwin King

ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 496–507 · Published 6 August 2026 · DOI 10.1145/3770855.3817617

Read the free full textCite this paper

In brief

“LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition” is a 2026 conference paper by Yanyu Chen and 10 co-authors, published in ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD). OpenAlex counts 0 citations as of 15 November 2026. It is open access under a CC BY licence: the full text is free to read at ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD).

Citations
0
References
34
Authors
11

Abstract

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.

Cite this paper

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

APA 7

Chen, Y., Jiang, J., Yu, D., Wu, Z., Liu, J., Han, J., Guo, X., Qi, J., Li, Y., Zhang, Y., & King, I. (2026). LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition. ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 496–507. https://doi.org/10.1145/3770855.3817617

MLA 9

Chen, Yanyu, et al. “LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition.” ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2026, pp. 496–507. https://doi.org/10.1145/3770855.3817617.

Chicago

Chen, Yanyu, Jiyue Jiang, Dianzhi Yu, Zheng Wu, Jiahong Liu, Jiaming Han, Xiao Guo, et al. 2026. “LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition.” ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD): 496–507. https://doi.org/10.1145/3770855.3817617.

Harvard

Chen, Y. et al. (2026) ‘LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition’, ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), pp. 496–507. doi:10.1145/3770855.3817617.

IEEE

Y. Chen et al., “LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition,” ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), pp. 496–507, 2026, doi: 10.1145/3770855.3817617.

BibTeX

@inproceedings{chen2026mining,
  title     = {{LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition}},
  author    = {Chen, Yanyu and Jiang, Jiyue and Yu, Dianzhi and Wu, Zheng and Liu, Jiahong and Han, Jiaming and Guo, Xiao and Qi, Jinhu and Li, Yu and Zhang, Yifei and King, Irwin},
  booktitle = {ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD)},
  year      = {2026},
  pages     = {496--507},
  doi       = {10.1145/3770855.3817617},
  url       = {https://doi.org/10.1145/3770855.3817617}
}

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 model reasoning
  • self-alignment
  • reward decomposition
  • label noise

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?

Chen, Y., Jiang, J., Yu, D., Wu, Z., Liu, J., Han, J., Guo, X., Qi, J., Li, Y., Zhang, Y., & King, I. (2026). LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition. ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 496–507. https://doi.org/10.1145/3770855.3817617

How many times has this paper been cited?

0 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 ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD).

Who wrote this paper?

Yanyu Chen, Jiyue Jiang, Dianzhi Yu, Zheng Wu, Jiahong Liu, Jiaming Han, Xiao Guo, Jinhu Qi, Yu Li, Yifei Zhang, Irwin King.

What is the DOI of this paper?

10.1145/3770855.3817617. The permanent link is https://doi.org/10.1145/3770855.3817617.

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.