When Two LLMs Debate, Both Think They'll Win

• URL: http://arxiv.org/abs/2505.19184v3
• Date: Mon, 09 Jun 2025 17:54:25 GMT
• Title: When Two LLMs Debate, Both Think They'll Win
• Authors: Pradyumna Shyama Prasad, Minh Nhat Nguyen,
• Abstract summary: We evaluate Large Language Models (LLMs) in a dynamic, adversarial debate setting.<n>We organized 60 three-round policy debates among ten state-of-the-art LLMs.<n>We observed five concerning patterns.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Can LLMs accurately adjust their confidence when facing opposition? Building on previous studies measuring calibration on static fact-based question-answering tasks, we evaluate Large Language Models (LLMs) in a dynamic, adversarial debate setting, uniquely combining two realistic factors: (a) a multi-turn format requiring models to update beliefs as new information emerges, and (b) a zero-sum structure to control for task-related uncertainty, since mutual high-confidence claims imply systematic overconfidence. We organized 60 three-round policy debates among ten state-of-the-art LLMs, with models privately rating their confidence (0-100) in winning after each round. We observed five concerning patterns: (1) Systematic overconfidence: models began debates with average initial confidence of 72.9% vs. a rational 50% baseline. (2) Confidence escalation: rather than reducing confidence as debates progressed, debaters increased their win probabilities, averaging 83% by the final round. (3) Mutual overestimation: in 61.7% of debates, both sides simultaneously claimed >=75% probability of victory, a logical impossibility. (4) Persistent self-debate bias: models debating identical copies increased confidence from 64.1% to 75.2%; even when explicitly informed their chance of winning was exactly 50%, confidence still rose (from 50.0% to 57.1%). (5) Misaligned private reasoning: models' private scratchpad thoughts sometimes differed from their public confidence ratings, raising concerns about faithfulness of chain-of-thought reasoning. These results suggest LLMs lack the ability to accurately self-assess or update their beliefs in dynamic, multi-turn tasks; a major concern as LLMs are now increasingly deployed without careful review in assistant and agentic roles. Code for our experiments is available at https://github.com/pradyuprasad/llms_overconfidence

Related papers

• How Overconfidence in Initial Choices and Underconfidence Under Criticism Modulate Change of Mind in Large Language Models [28.62988505317048]
  Large language models (LLMs) exhibit strikingly conflicting behaviors.<n>LLMs can appear steadfastly overconfident in their initial answers whilst being prone to excessive doubt when challenged.<n>We show that LLMs exhibit a pronounced choice-supportive bias that reinforces and boosts their estimate of confidence in their answer.
  arXiv  Detail & Related papers  (2025-07-03T18:57:43Z)
• ConfQA: Answer Only If You Are Confident [49.34040922485979]
  We present a fine-tuning strategy that we call ConfQA, which can reduce hallucination rate from 20-40% to under 5% across multiple factuality benchmarks.<n>We introduce a dampening prompt "answer only if you are confident" to explicitly guide the behavior, without which hallucination remains high as 15%-25%.<n>We propose the Dual Neural Knowledge framework, which seamlessly select between internally parameterized neural knowledge and externally recorded symbolic knowledge.
  arXiv  Detail & Related papers  (2025-06-08T22:51:46Z)
• When Persuasion Overrides Truth in Multi-Agent LLM Debates: Introducing a Confidence-Weighted Persuasion Override Rate (CW-POR) [0.46040036610482665]
  In many real-world scenarios, a single Large Language Model (LLM) may encounter contradictory claims-some accurate, others forcefully incorrect-and must judge which is true.<n>We investigate this risk in a single-turn, multi-agent debate framework: one LLM-based agent provides a factual answer from TruthfulQA, another vigorously defends a falsehood, and the same architecture serves as judge.<n>We introduce the Confidence-Weighted Persuasion Rate (CW-POR), which captures not only how often the judge is deceived but also how strongly it believes the incorrect choice.
  arXiv  Detail & Related papers  (2025-04-01T02:45:02Z)
• SteerConf: Steering LLMs for Confidence Elicitation [11.872504642312705]
  Large Language Models (LLMs) exhibit impressive performance across diverse domains but often suffer from overconfidence.<n>We propose SteerConf, a novel framework that systematically steers LLMs' confidence scores to improve their calibration and reliability.
  arXiv  Detail & Related papers  (2025-03-04T18:40:49Z)
• Language Models Prefer What They Know: Relative Confidence Estimation via Confidence Preferences [62.52739672949452]
  Language models (LMs) should provide reliable confidence estimates to help users detect mistakes in their outputs and defer to human experts when necessary.<n>We propose relative confidence estimation, where we match up questions against each other and ask the model to make relative judgments of confidence.<n>Treating each question as a "player" in a series of matchups against other questions and the model's preferences as match outcomes, we can use rank aggregation methods like Elo rating and Bradley-Terry to translate the model's confidence preferences into confidence scores.
  arXiv  Detail & Related papers  (2025-02-03T07:43:27Z)
• Confidence in the Reasoning of Large Language Models [0.0]
  Confidence is measured in terms of persistence in keeping their answer when prompted to reconsider.<n> Confidence is only partially explained by the underlying token-level probability.
  arXiv  Detail & Related papers  (2024-12-19T10:04:29Z)
• DebUnc: Improving Large Language Model Agent Communication With Uncertainty Metrics [52.242449026151846]
  Multi-agent debates have been introduced to improve the accuracy of Large Language Models (LLMs)<n>We propose DebUnc, a debate framework that uses uncertainty metrics to assess agent confidence.
  arXiv  Detail & Related papers  (2024-07-08T22:15:01Z)
• Factual Confidence of LLMs: on Reliability and Robustness of Current Estimators [6.403926452181712]
  Large Language Models (LLMs) tend to be unreliable in the factuality of their answers. We present a survey and empirical comparison of estimators of factual confidence. Our experiments indicate that trained hidden-state probes provide the most reliable confidence estimates.
  arXiv  Detail & Related papers  (2024-06-19T10:11:37Z)
• Think Twice Before Trusting: Self-Detection for Large Language Models through Comprehensive Answer Reflection [90.71323430635593]
  We propose a novel self-detection paradigm that considers the comprehensive answer space beyond LLM-generated answers. Building upon this paradigm, we introduce a two-step framework, which firstly instructs LLM to reflect and provide justifications for each candidate answer. This framework can be seamlessly integrated with existing approaches for superior self-detection.
  arXiv  Detail & Related papers  (2024-03-15T02:38:26Z)
• Reconfidencing LLMs from the Grouping Loss Perspective [56.801251926946485]
  Large Language Models (LLMs) are susceptible to generating hallucinated answers in a confident tone. Recent findings show that controlling uncertainty must go beyond calibration. We construct a new evaluation dataset derived from a knowledge base to assess confidence scores given to answers of Mistral and LLaMA.
  arXiv  Detail & Related papers  (2024-02-07T15:40:22Z)
• Llamas Know What GPTs Don't Show: Surrogate Models for Confidence Estimation [70.27452774899189]
  Large language models (LLMs) should signal low confidence on examples where they are incorrect, instead of misleading the user. As of November 2023, state-of-the-art LLMs do not provide access to these probabilities. Our best method composing linguistic confidences and surrogate model probabilities gives state-of-the-art confidence estimates on all 12 datasets.
  arXiv  Detail & Related papers  (2023-11-15T11:27:44Z)
• Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs [60.61002524947733]
  Previous confidence elicitation methods rely on white-box access to internal model information or model fine-tuning. This leads to a growing need to explore the untapped area of black-box approaches for uncertainty estimation. We define a systematic framework with three components: prompting strategies for eliciting verbalized confidence, sampling methods for generating multiple responses, and aggregation techniques for computing consistency.
  arXiv  Detail & Related papers  (2023-06-22T17:31:44Z)

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.