Candidate supply and answer selection shape the value of LLM judging in multi-agent systems

• URL: http://arxiv.org/abs/2608.25937v2
• Date: Sun, 30 Aug 2026 17:09:24 GMT
• Title: Candidate supply and answer selection shape the value of LLM judging in multi-agent systems
• Abstract summary: A correct answer is often already present among the generated candidates, but the system can still converge on and report a wrong answer.<n>We study two questions: (1) when an LLM judge provides effective selection pressure by supplying a signal of answer correctness for candidates generated in a multi-agent system, and (2) when using that signal improves the reported answer.
• Score: 4.178517265499193
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multi-agent systems (MAS) sometimes already have the potential to answer correctly, but still report a wrong answer. Explaining this outcome is difficult because generation, communication and final answer-selection rules usually change simultaneously. We conceptualize multi-agent reasoning as an evolutionary pipeline of candidate generation, peer communication and terminal selection, wherein consensus without quality control can exhibit patterns of memetic drift. We study two questions: (1) when an LLM judge provides effective selection pressure by supplying a signal of answer correctness for candidates generated in a multi-agent system, and (2) when using that signal improves the reported answer. To map judge reliability, we analysed 15,336 questions from MMLU-Pro, GPQA, MedXpertQA and MuSR, with Humanity's Last Exam analysed separately. To test these rules, we replayed 81,390 fixed candidate pools drawn from 16,278 questions across five benchmarks. We report three findings. (1) A correct answer is often already present among the generated candidates, but the system can still converge on and report a wrong answer. (2) Judge reliability is not a fixed trait of the model, but varies with the task, the generator and how rare the correct answer is. (3) Combining answer frequency with the judge's evaluation changed only the final answer-selection rule and raised accuracy from 63.82% to 70.82-70.95%, primarily by rescuing correct answers that were outnumbered by popular errors. In the systems studied here, the value of generating more candidates depends on whether those extra samples make correct answers present, frequent or recognisable. By isolating generation, recognition and selection, these findings establish a diagnostic basis for designing multi-agent architectures that protect generated correct answers from being lost.

Related papers

• A decodability criterion predicts when hidden-state selection beats majority voting in large language models [10.860308169044679]
  We propose a dynamic selection combiner that trains a linear gate on the answer-token hidden state and selects the highest-scoring candidate.<n>Its main contribution is decodability, a leakage-free measure of how well the gate ranks a question's correct candidates.
  arXiv  Detail & Related papers  (2026-08-17T21:02:10Z)
• When Many Answers Are Valid, Voting Fails: Symbolic Verification for Best-of-K Causal Reasoning in LLMs [42.32694162421078]
  We introduce CALVER, a training-free symbolic verifier that scores structured traces against Pearl's causal criteria.<n>On CLEAR find-one-valid queries that admit multiple graph-valid answers, CALVER reaches 42.1% where plurality, a reward model, an LLM judge, and model confidence remain near 30%.
  arXiv  Detail & Related papers  (2026-08-04T11:45:46Z)
• Boosting Self-Consistency with Ranking [56.38798757709555]
  Self-consistency improves large language models by sampling multiple reasoning paths and selecting the most frequent answer.<n>We address this limitation with Ranking-Improved Self-Consistency (RISC), which reformulates answer selection in self-consistency as a ranking problem.
  arXiv  Detail & Related papers  (2026-06-03T16:12:30Z)
• Reaching Beyond the Mode: RL for Distributional Reasoning in Language Models [78.68818219506313]
  This paper describes a multi-answer reinforcement learning approach for training LMs to perform distributional reasoning over multiple answers during inference.<n>Across question-answering, medical diagnostic, and coding benchmarks, we observe improved diversity, coverage, and set-level calibration scores compared to single answer trained baselines.
  arXiv  Detail & Related papers  (2026-03-25T22:20:25Z)
• Answer Matching Outperforms Multiple Choice for Language Model Evaluation [35.90520208701438]
  We show multiple choice questions from popular benchmarks can often be answered without even seeing the question.<n>We consider generative evaluation via what we call answer matching.
  arXiv  Detail & Related papers  (2025-07-03T17:59:02Z)
• Right Answer, Wrong Score: Uncovering the Inconsistencies of LLM Evaluation in Multiple-Choice Question Answering [78.89231943329885]
  Multiple-Choice Question Answering (MCQA) is widely used to evaluate Large Language Models (LLMs)<n>We show that multiple factors can significantly impact the reported performance of LLMs.<n>We analyze whether existing answer extraction methods are aligned with human judgment.
  arXiv  Detail & Related papers  (2025-03-19T08:45:03Z)
• Multimodal Reranking for Knowledge-Intensive Visual Question Answering [77.24401833951096]
  We introduce a multi-modal reranker to improve the ranking quality of knowledge candidates for answer generation. Experiments on OK-VQA and A-OKVQA show that multi-modal reranker from distant supervision provides consistent improvements.
  arXiv  Detail & Related papers  (2024-07-17T02:58:52Z)
• Answer Generation for Retrieval-based Question Answering Systems [80.28727681633096]
  We train a sequence to sequence transformer model to generate an answer from a candidate set. Our tests on three English AS2 datasets show improvement up to 32 absolute points in accuracy over the state of the art.
  arXiv  Detail & Related papers  (2021-06-02T05:45:49Z)
• MS-Ranker: Accumulating Evidence from Potentially Correct Candidates for Answer Selection [59.95429407899612]
  We propose a novel reinforcement learning based multi-step ranking model, named MS-Ranker. We explicitly consider the potential correctness of candidates and update the evidence with a gating mechanism. Our model significantly outperforms existing methods that do not rely on external resources.
  arXiv  Detail & Related papers  (2020-10-10T10:36:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.