Debate, Deliberate, Decide (D3): A Cost-Aware Adversarial Framework for Reliable and Interpretable LLM Evaluation

• URL: http://arxiv.org/abs/2410.04663v3
• Date: Sat, 11 Oct 2025 20:38:04 GMT
• Title: Debate, Deliberate, Decide (D3): A Cost-Aware Adversarial Framework for Reliable and Interpretable LLM Evaluation
• Authors: Chaithanya Bandi, Abir Harrasse,
• Abstract summary: We present Debate, Deliberate, Decide (D3), a cost-aware, adversarial multi-agent framework that orchestrates structured debate among role-specialized agents.<n>We develop a probabilistic model of score gaps that characterizes reliability and convergence under iterative debate.<n>We show state-of-the-art agreement with human judgments, reduced positional and verbosity biases via anonymization, and a favorable cost-accuracy frontier.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The evaluation of Large Language Models (LLMs) remains challenging due to inconsistency, bias, and the absence of transparent decision criteria in automated judging. We present Debate, Deliberate, Decide (D3), a cost-aware, adversarial multi-agent framework that orchestrates structured debate among role-specialized agents (advocates, a judge, and an optional jury) to produce reliable and interpretable evaluations. D3 instantiates two complementary protocols: (1) Multi-Advocate One-Round Evaluation (MORE), which elicits k parallel defenses per answer to amplify signal via diverse advocacy, and (2) Single-Advocate Multi-Round Evaluation (SAMRE) with budgeted stopping, which iteratively refines arguments under an explicit token budget and convergence checks. We develop a probabilistic model of score gaps that (i) characterizes reliability and convergence under iterative debate and (ii) explains the separation gains from parallel advocacy. Under mild assumptions, the posterior distribution of the round-r gap concentrates around the true difference and the probability of mis-ranking vanishes; moreover, aggregating across k advocates provably increases expected score separation. We complement theory with a rigorous experimental suite across MT-Bench, AlignBench, and AUTO-J, showing state-of-the-art agreement with human judgments (accuracy and Cohen's kappa), reduced positional and verbosity biases via anonymization and role diversification, and a favorable cost-accuracy frontier enabled by budgeted stopping. Ablations and qualitative analyses isolate the contributions of debate, aggregation, and anonymity. Together, these results establish D3 as a principled, practical recipe for reliable, interpretable, and cost-aware LLM evaluation.

Related papers

• Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution [79.98699884805636]
  Reasoning Execution by Multiple Listeners (REMUL) is a multi-party reinforcement learning approach.<n>REMUL builds on the hypothesis that reasoning traces which other parties can follow will be more faithful.<n>Speakers are rewarded for producing reasoning that is clear to listeners.
  arXiv  Detail & Related papers  (2026-02-18T02:55:55Z)
• Multimodal Fact-Level Attribution for Verifiable Reasoning [80.60864342985748]
  Multimodal large language models (MLLMs) are increasingly used for real-world tasks involving multi-step reasoning and long-form generation.<n>Existing multimodal grounding benchmarks and evaluation methods fail to assess attribution in complex multimodal reasoning.<n>We introduce MuRGAt, a benchmark for evaluating fact-level multimodal attribution in settings that require reasoning beyond direct observation.
  arXiv  Detail & Related papers  (2026-02-12T03:10:02Z)
• A Comprehensive Evaluation of LLM Reasoning: From Single-Model to Multi-Agent Paradigms [20.241519889633285]
  Large Language Models (LLMs) are increasingly deployed as reasoning systems, where reasoning paradigms play a critical role.<n>We conduct a comprehensive and unified evaluation of reasoning paradigms, spanning direct single-model generation, CoT-augmented single-model reasoning, and representative MAS.<n>We introduce MIMeBench, a new open-ended benchmark that targets two foundational yet underexplored semantic capabilities.
  arXiv  Detail & Related papers  (2026-01-19T17:23:45Z)
• The Stability Trap: Evaluating the Reliability of LLM-Based Instruction Adherence Auditing [1.5954459915735735]
  This study asks: To what extent does the instruction type of an Application Under Test (AUT) influence the stability of judge evaluations?<n>We introduce the Scoped Instruction Decomposition Framework to classify AUT instructions into Objective and Subjective types, isolating the factors that drive judge instability.<n>Our results reveal a Stability Trap'' characterized by a divergence between Verdict Stability and Reasoning Stability.
  arXiv  Detail & Related papers  (2026-01-16T21:15:13Z)
• Adversarial Yet Cooperative: Multi-Perspective Reasoning in Retrieved-Augmented Language Models [72.4149653187766]
  We propose a Reasoner-Verifier framework named Adrialversa Reasoning RAG (ARR)<n>The Reasoner and Verifier engage in reasoning on retrieved evidence and critiquing each other's logic while being guided by process-aware advantage.<n> Experiments on multiple benchmarks demonstrate the effectiveness of our method.
  arXiv  Detail & Related papers  (2026-01-08T06:57:03Z)
• Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following [99.20581206115979]
  Multi-Crit is a benchmark for evaluating multimodal judges on their capacity to follow pluralistic criteria and produce reliable criterion-level judgments.<n> Comprehensive analysis of 25 LMMs reveals that 1) proprietary models still struggle to maintain consistent adherence to pluralistic criteria--especially in open-ended evaluation; 2) open-source models lag further behind in flexibly following diverse criteria; and 3) critic fine-tuning with holistic judgment signals enhances visual grounding but fails to generalize to pluralistic criterion-level judgment.
  arXiv  Detail & Related papers  (2025-11-26T18:35:17Z)
• TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them [58.04324690859212]
  Large Language Models (LLMs) as automated evaluators (LLM-as-a-judge) has revealed critical inconsistencies in current evaluation frameworks.<n>We identify two fundamental types of inconsistencies: Score-Comparison Inconsistency and Pairwise Transitivity Inconsistency.<n>We propose TrustJudge, a probabilistic framework that addresses these limitations through two key innovations.
  arXiv  Detail & Related papers  (2025-09-25T13:04:29Z)
• CMR-SPB: Cross-Modal Multi-Hop Reasoning over Text, Image, and Speech with Path Balance [10.843417240658992]
  Cross-modal multi-hop reasoning (CMR) is a valuable yet underexplored capability of multimodal large language models (MLLMs)<n>We argue that existing benchmarks for evaluating this ability have critical shortcomings.<n>We introduce a novel benchmark -- Cross-Modal Multi-Hop Reasoning over Text, Image and Speech with Path Balance (CMR-SPB)
  arXiv  Detail & Related papers  (2025-08-22T08:17:31Z)
• Debate, Reflect, and Distill: Multi-Agent Feedback with Tree-Structured Preference Optimization for Efficient Language Model Enhancement [43.532921045069365]
  Large Language Models (LLMs) continue to set new standards in knowledge-intensive and complex reasoning tasks.<n>Current techniques, such as static knowledge distillation, resource-intensive reinforcement learning from human feedback, or limited self-reflection to yield substantial and lasting performance gains.<n>We present a novel Reflect and Debate (D&R) framework that orchestrates multi-turn debates between smaller models and stronger teacher models, eliciting actionable feedback.
  arXiv  Detail & Related papers  (2025-06-04T03:52:20Z)
• KORGym: A Dynamic Game Platform for LLM Reasoning Evaluation [78.96590724864606]
  We introduce the Knowledge Orthogonal Reasoning Gymnasium (KORGym), a dynamic evaluation platform inspired by KOR-Bench and Gymnasium.<n>KORGym offers over fifty games in either textual or visual formats and supports interactive, multi-turn assessments with reinforcement learning scenarios.
  arXiv  Detail & Related papers  (2025-05-20T16:06:32Z)
• Ensemble Bayesian Inference: Leveraging Small Language Models to Achieve LLM-level Accuracy in Profile Matching Tasks [0.0]
  This study explores the potential of small language model(SLM) ensembles to achieve accuracy comparable to proprietary large language models (LLMs) We propose Ensemble Bayesian Inference (EBI), a novel approach that applies Bayesian estimation to combine judgments from multiple SLMs.
  arXiv  Detail & Related papers  (2025-04-24T15:55:10Z)
• Debate-Feedback: A Multi-Agent Framework for Efficient Legal Judgment Prediction [7.196065223124077]
  We propose a novel legal judgment prediction model based on the Debate-Feedback architecture. Unlike traditional methods, our model achieves significant improvements in efficiency by minimizing the need for large historical datasets.
  arXiv  Detail & Related papers  (2025-04-07T09:34:14Z)
• Multi-Agent LLM Judge: automatic personalized LLM judge design for evaluating natural language generation applications [0.0]
  Large Language Models (LLMs) have demonstrated impressive performance across diverse domains, yet they still encounter challenges such as insufficient domain-specific knowledge, biases, and hallucinations. Traditional evaluation methods, which rely on word overlap or text embeddings, are inadequate for capturing the nuanced semantic information necessary to evaluate dynamic, open-ended text generation. We propose a novel dynamic multi-agent system that automatically designs personalized LLM judges for various natural language generation applications.
  arXiv  Detail & Related papers  (2025-04-01T09:36:56Z)
• Revisiting Benchmark and Assessment: An Agent-based Exploratory Dynamic Evaluation Framework for LLMs [29.72874725703848]
  We introduce two concepts: Benchmark+, which extends traditional question-answer benchmark into a more flexible "strategy-criterion" format; and Assessment+, which enhances the interaction process. We propose an agent-based evaluation framework called TestAgent, which implements these concepts through retrieval augmented generation and reinforcement learning.
  arXiv  Detail & Related papers  (2024-10-15T11:20:42Z)
• Thinking Fair and Slow: On the Efficacy of Structured Prompts for Debiasing Language Models [14.405446719317291]
  Existing debiasing techniques are typically training-based or require access to the model's internals and output distributions. We evaluate a comprehensive end-user-focused iterative framework of debiasing that applies System 2 thinking processes for prompts to induce logical, reflective, and critical text generation.
  arXiv  Detail & Related papers  (2024-05-16T20:27:58Z)
• DEBATE: Devil's Advocate-Based Assessment and Text Evaluation [6.2689399557794525]
  We propose DEBATE, an NLG evaluation framework based on multi-agent scoring system. Within the framework, one agent is instructed to criticize other agents' arguments. We show that the extensiveness of debates among agents and the persona of an agent can influence the performance of evaluators.
  arXiv  Detail & Related papers  (2024-05-16T09:41:12Z)
• Large Multimodal Agents: A Survey [78.81459893884737]
  Large language models (LLMs) have achieved superior performance in powering text-based AI agents. There is an emerging research trend focused on extending these LLM-powered AI agents into the multimodal domain. This review aims to provide valuable insights and guidelines for future research in this rapidly evolving field.
  arXiv  Detail & Related papers  (2024-02-23T06:04:23Z)
• LLM Comparator: Visual Analytics for Side-by-Side Evaluation of Large Language Models [31.426274932333264]
  We present Comparator, a novel visual analytics tool for interactively analyzing results from automatic side-by-side evaluation. The tool supports interactive for users to understand when and why a model performs better or worse than a baseline model.
  arXiv  Detail & Related papers  (2024-02-16T09:14:49Z)
• Can Large Language Models be Trusted for Evaluation? Scalable Meta-Evaluation of LLMs as Evaluators via Agent Debate [74.06294042304415]
  We propose ScaleEval, an agent-debate-assisted meta-evaluation framework. We release the code for our framework, which is publicly available on GitHub.
  arXiv  Detail & Related papers  (2024-01-30T07:03:32Z)
• DCR-Consistency: Divide-Conquer-Reasoning for Consistency Evaluation and Improvement of Large Language Models [4.953092503184905]
  This work proposes DCR, an automated framework for evaluating and improving the consistency of Large Language Models (LLMs) generated texts. We introduce an automatic metric converter (AMC) that translates the output from DCE into an interpretable numeric score. Our approach also substantially reduces nearly 90% of output inconsistencies, showing promise for effective hallucination mitigation.
  arXiv  Detail & Related papers  (2024-01-04T08:34:16Z)
• Answering Causal Queries at Layer 3 with DiscoSCMs-Embracing Heterogeneity [0.0]
  This paper advocates for the Distribution-consistency Structural Causal Models (DiscoSCM) framework as a pioneering approach to counterfactual inference.
  arXiv  Detail & Related papers  (2023-09-17T17:01:05Z)
• ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate [57.71597869337909]
  We build a multi-agent referee team called ChatEval to autonomously discuss and evaluate the quality of generated responses from different models. Our analysis shows that ChatEval transcends mere textual scoring, offering a human-mimicking evaluation process for reliable assessments.
  arXiv  Detail & Related papers  (2023-08-14T15:13:04Z)
• Unlocking the Potential of User Feedback: Leveraging Large Language Model as User Simulator to Enhance Dialogue System [65.93577256431125]
  We propose an alternative approach called User-Guided Response Optimization (UGRO) to combine it with a smaller task-oriented dialogue model. This approach uses LLM as annotation-free user simulator to assess dialogue responses, combining them with smaller fine-tuned end-to-end TOD models. Our approach outperforms previous state-of-the-art (SOTA) results.
  arXiv  Detail & Related papers  (2023-06-16T13:04:56Z)
• Rethinking the Evaluation for Conversational Recommendation in the Era of Large Language Models [115.7508325840751]
  The recent success of large language models (LLMs) has shown great potential to develop more powerful conversational recommender systems (CRSs) In this paper, we embark on an investigation into the utilization of ChatGPT for conversational recommendation, revealing the inadequacy of the existing evaluation protocol. We propose an interactive Evaluation approach based on LLMs named iEvaLM that harnesses LLM-based user simulators.
  arXiv  Detail & Related papers  (2023-05-22T15:12:43Z)
• Perspectives on Large Language Models for Relevance Judgment [56.935731584323996]
  Large language models (LLMs) claim that they can assist with relevance judgments. It is not clear whether automated judgments can reliably be used in evaluations of retrieval systems.
  arXiv  Detail & Related papers  (2023-04-13T13:08:38Z)
• Exploring the Trade-off between Plausibility, Change Intensity and Adversarial Power in Counterfactual Explanations using Multi-objective Optimization [73.89239820192894]
  We argue that automated counterfactual generation should regard several aspects of the produced adversarial instances. We present a novel framework for the generation of counterfactual examples.
  arXiv  Detail & Related papers  (2022-05-20T15:02:53Z)

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.