Abstract
Multi-path reasoning methods such as self-consistency (SC) sample $K$ reasoning paths and choose the most frequent answer. However, their gains quickly plateau as $K$ increases, and existing methods do not predict when this saturation will occur. We formalize multi-path LLM reasoning as a diversity combining problem from wireless communications: each path is a noisy channel observation, and the pairwise correlation of path correctness caps the design-effect effective sample size of the vote at a finite ceiling. Generalized least squares (GLS) analysis shows that, under exchangeability, the optimal symmetric linear combiner of latent embeddings is uniform, supporting majority vote as the natural default in standard SC while leaving room for weighting or pruning under heterogeneous prompt-template branches. Across 5 models and 12 benchmarks, prompt-template diversity reduces path correlation in $55$ of $57$ valid cells, with the strongest effect on open-ended QA. We derive an Adaptive-K rule that uses a four-path pilot to select $K^*$, retaining $96$--$103\%$ of MV@$K{=}32$ accuracy across Math, QA, and NLU.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Yu, G., Zhang, L., Wang, Q., Wang, X., Li, M., Ji, S., Liu, R. P., & Piccardi, M. (2026). Diversity Combining for Multi-Path LLM Reasoning. https://omanscience.com/en/articles/diversity-combining-for-multi-path-llm-reasoning
MLA 9
Yu, Guangsheng, et al. "Diversity Combining for Multi-Path LLM Reasoning." https://omanscience.com/en/articles/diversity-combining-for-multi-path-llm-reasoning.
Chicago (author–date)
Yu, Guangsheng, Litianyi Zhang, Qin Wang, Xu Wang, Mingyuan Li, Shaoxiong Ji, Ren Ping Liu, and Massimo Piccardi. 2026. "Diversity Combining for Multi-Path LLM Reasoning." https://omanscience.com/en/articles/diversity-combining-for-multi-path-llm-reasoning.
Harvard
Yu, G., Zhang, L., Wang, Q., Wang, X., Li, M., Ji, S., Liu, R. P. and Piccardi, M. (2026) 'Diversity Combining for Multi-Path LLM Reasoning', Available at: https://omanscience.com/en/articles/diversity-combining-for-multi-path-llm-reasoning.
Vancouver
Yu G, Zhang L, Wang Q, Wang X, Li M, Ji S, et al. Diversity Combining for Multi-Path LLM Reasoning. https://omanscience.com/en/articles/diversity-combining-for-multi-path-llm-reasoning
IEEE
G. Yu, L. Zhang, Q. Wang, X. Wang, M. Li, S. Ji, R. P. Liu, and M. Piccardi, "Diversity Combining for Multi-Path LLM Reasoning," https://omanscience.com/en/articles/diversity-combining-for-multi-path-llm-reasoning.