analiza aiqbench
This research addresses the "degeneration of thought" observed when single LLMs repeatedly reflect on their own errors by introducing a multi-agent debate framework. By leveraging diverse personas to generate reflections, the approach achieves superior performance on reasoning benchmarks like HotPot QA and HumanEval compared to self-reflection methods. The key takeaway for practitioners is that introducing agent diversity can effectively mitigate the stagnation of iterative self-correction in complex reasoning tasks.
MAR:Multi-Agent Reflexion Improves Reasoning Abilities in LLMs
arXiv:2512.20845v2 Announce Type: replace Abstract: LLMs have shown the capacity to improve their performance on reasoning tasks through reflecting on their mistakes, and acting with these reflections in mind. However, continual reflections of the same LLM onto itself exhibit degeneration of thought, where the LLM continues to repeat the same errors again and again even with the knowledge that its wrong. To address this problem, we instead introduce multi-agent with multi-persona debators as the method to generate reflections. Through out extensive experimentation, we've found that the leads to better diversity of in the reflections generated by the llm agent. We demonstrate an accuracy of 47% EM HotPot QA (question answering) and 82.7% on HumanEval (programming), both performances surpassing reflection with a single llm.
Abstract:LLMs have shown the capacity to improve their performance on reasoning tasks through reflecting on their mistakes, and acting with these reflections in mind. However, continual reflections of the same LLM onto itself exhibit degeneration of thought, where the LLM continues to repeat the same errors again and again even with the knowledge that its wrong. To address this problem, we instead introduce multi-agent with multi-persona debators as the method to generate reflections. Through out extensive experimentation, we've found that the leads to better diversity of in the reflections generated by the llm agent. We demonstrate an accuracy of 47% EM HotPot QA (question answering) and 82.7% on HumanEval (programming), both performances surpassing reflection with a single llm.
Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
arXiv:2512.20845 [cs.AI]
(or arXiv:2512.20845v2 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2512.20845 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 213 cuvinte