mlsanityData to 7 Oct 2026

Unanimity Without Persuasion: A Single Round of Debate Erases the Disagreement That Verification Needs

arXiv 2609.26145 ↗First posted 10 Aug 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

A debate panel can become unanimous without becoming more correct. This is dangerous for downstream safeguards: a substituted verification ballot can change only narrow-margin votes, while richer arbiters lose disagreement as a natural targeting signal. We show that one debate round can erase that resource without requiring persuasion. Tracking a heterogeneous 7-judge panel through a blind round and three debate rounds on 600 code-correctness candidates, unanimity on a fixed cohort jumps from 39.5% to 95.2% in round 1 (93.1% of the total collapse), while accuracy moves by less than one point and 96.3% of verdict flips follow the displayed peer majority. An execution-based verification ballot corrects 8 of 2,037 pre-debate candidate-substitution instances but changes zero in every later round; by round 3 every wrong decision is unanimous, erasing dissent that had flagged two-thirds of the panel's errors. Identical-cohort controls explain why: no-peer reconsideration reproduces 79.6% of the collapse, real labels without reasoning reproduce 91.5%, and random labels steer flips toward whatever they display; the full-debate condition adds 4.3 percentage points over labels only (clustered 95% CI 0.7--8.1). The one-round collapse reproduces in two additional real runs and two fake-label seeds, remains under panel sizes 3--7, and appears in MATH-500. Parse failures concentrate on contested candidates (p<0.001), making attrition non-ignorable. The design implication is operational: verify before any second-pass evaluation or peer exposure, and never treat post-debate unanimity as independent evidence of reliability.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 19 of its 26 references that are in the map. First posted on arXiv 10 Aug 2026.

Builds on

19 of 26 references.

References · 19

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  4. 2021Program Synthesis with Large Language Modelscited by 3,509
  5. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  6. 2023Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debatecited by 995
  7. 2023Towards Understanding Sycophancy in Language Modelscited by 988
  8. 2017Selective Classification for Deep Neural Networkscited by 799
  9. 2023ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debatecited by 724
  10. 2018AI safety via debatecited by 384
  11. 2024Debating with More Persuasive LLMs Leads to More Truthful Answerscited by 255
  12. 2023Should we be going MAD? A Look at Multi-Agent Debate Strategies for LLMscited by 103
7 more
  1. 2025Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models?cited by 74
  2. 2025Correlated Errors in Large Language Modelscited by 42
  3. 2024Conformity in Large Language Modelscited by 30
  4. 2025Can LLM Agents Really Debate? A Controlled Study of Multi-Agent Debate in Logical Reasoningcited by 27
  5. 2026Nine Judges, Two Effective Votes: Correlated Errors Undermine LLM Evaluation Panelscited by 21
  6. 2026Demystifying Multi-Agent Debate: The Role of Confidence and Diversitycited by 17
  7. 2026Minority Sentinel: When to Overturn Majority Voting in Multi-Agent LLM Debatescited by 4

Built on by

Not cited yet.