Multi-Agent Post-Co-Training of Large Language Models via Reinforcement Learning
Frontiers in Applied Physics and Mathematics · 10.03.2026 · 3 (2) · 10-16
DOI: 10.71465/fapm716 ↗Kurz gefasst
Fachartikel in der Zeitschrift Frontiers in Applied Physics and Mathematics, online erschienen am 10.03.2026. Verfasst von James L. Carter und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This study introduces MAPoRL2, a post-training framework that enhances collaborative LLM performance through multi-agent reinforcement learning and structured discussion. Multiple LLM agents independently generate candidate solutions, engage in iterative discussion rounds, and are jointly optimized using verifier-based rewards that assess both correctness and corrective reasoning. Experiments across 5 reasoning and generation benchmarks with 4,500 training samples demonstrate improvements of 18.9% in answer accuracy and 22.4% in correction efficiency over single-agent post-training, highlighting the effectiveness of discussion-aware RL signals.
Zitieren
APA
Carter, J. L., Liu, Y., & Lee, T. K. (2026). Multi-Agent Post-Co-Training of Large Language Models via Reinforcement Learning. Frontiers in Applied Physics and Mathematics, 3(2), 10-16. https://doi.org/10.71465/fapm716
Vancouver
Carter JL, Liu Y, Lee TK. Multi-Agent Post-Co-Training of Large Language Models via Reinforcement Learning. Frontiers in Applied Physics and Mathematics. 2026;3(2):10-16. doi:10.71465/fapm716
BibTeX
@article{Carter2026,
author = {Carter, James L. and Liu, Yuxuan and Lee, Thomas K.},
title = {Multi-Agent Post-Co-Training of Large Language Models via Reinforcement Learning},
journal = {Frontiers in Applied Physics and Mathematics},
year = {2026},
volume = {3},
number = {2},
pages = {10--16},
doi = {10.71465/fapm716}
}
EUVIMED