Abstract
Multi-agent collaboration lets large language models (LLMs) improve question answering through deliberation and feedback. Yet shared discussion couples correction with exposure to the same mistakes, which can erode the diversity needed for voting. Self-consistency offers sampling diversity without feedback, while single-pair Actor-Critic collaboration refines only one candidate. We introduce SEPAL, which assigns three private Actor-Critic teams to direct reasoning, evidence grounding, and verification. Role-specific training gives the teams different reasoning objectives beyond sampling variation. Each Critic guides revisions within its own team, preventing feedback from carrying errors across candidates. Once revision ends, majority voting combines only the final answers, keeping the reasoning histories separate until the decision. Across five open-weight backbones and five question-answering benchmarks, SEPAL improves mean accuracy by 1.81 percentage points over a matched single Actor-Critic pair, with improvements across all five backbones. Code is available at https://github.com/zhansan114514/SEPAL.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Ren, W., Zhang, Y., Li, H., Qi, Z., Zhang, H., & Wang, N. (2026). SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration. https://omanscience.com/en/articles/sepal-separated-expert-pairs-with-answer-level-fusion-for-reliable-llm-collaboration
MLA 9
Ren, Weijie, et al. "SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration." https://omanscience.com/en/articles/sepal-separated-expert-pairs-with-answer-level-fusion-for-reliable-llm-collaboration.
Chicago (author–date)
Ren, Weijie, Yanwen Zhang, Hao Li, Zhuolin Qi, Hengyi Zhang, and Naibo Wang. 2026. "SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration." https://omanscience.com/en/articles/sepal-separated-expert-pairs-with-answer-level-fusion-for-reliable-llm-collaboration.
Harvard
Ren, W., Zhang, Y., Li, H., Qi, Z., Zhang, H. and Wang, N. (2026) 'SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration', Available at: https://omanscience.com/en/articles/sepal-separated-expert-pairs-with-answer-level-fusion-for-reliable-llm-collaboration.
Vancouver
Ren W, Zhang Y, Li H, Qi Z, Zhang H, Wang N. SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration. https://omanscience.com/en/articles/sepal-separated-expert-pairs-with-answer-level-fusion-for-reliable-llm-collaboration
IEEE
W. Ren, Y. Zhang, H. Li, Z. Qi, H. Zhang, and N. Wang, "SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration," https://omanscience.com/en/articles/sepal-separated-expert-pairs-with-answer-level-fusion-for-reliable-llm-collaboration.