Metraipolis › archive › dc_0beea1bdecc6

Diversity Combining for Multi-Path LLM Reasoning — arXiv:2609.38829

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.38829
License (as archived)
CC0-1.0
Archived
2026-10-07T12:42:14.002Z
SHA-256 of the archived text
cf2bf1d752ee3ab474986c675e2d50e3acc30636c16d744967648166210ecd73
Stable URL of this page
https://metraipolis.com/documento/dc_0beea1bdecc6

How to cite

Diversity Combining for Multi-Path LLM Reasoning — arXiv:2609.38829 — arXiv (en), https://arxiv.org/abs/2609.38829 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0beea1bdecc6 (Metraipolis archive dc_0beea1bdecc6, sha256 cf2bf1d752ee3ab474986c675e2d50e3acc30636c16d744967648166210ecd73)
BibTeX
@misc{dc0beea1bdecc6,
  title = {Diversity Combining for Multi-Path LLM Reasoning — arXiv:2609.38829},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.38829},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0beea1bdecc6; sha256 cf2bf1d752ee3ab474986c675e2d50e3acc30636c16d744967648166210ecd73},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.38829. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.38829 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T12:14:13.694Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).