When to Trust Context: Self-Reflective Debates for Context Reliability

TL;DR

This paper introduces SR-DCR, a lightweight self-reflective debate framework that improves large language models' ability to assess context reliability, reducing hallucinations and factual errors in conflicting scenarios.

Contribution

The paper presents a novel debate-based approach combining token-level confidence and multi-agent debate to evaluate context trustworthiness in language models.

Findings

SR-DCR outperforms classical debate and confidence baselines.

It enhances robustness to misleading context.

Maintains accuracy on trustworthy inputs.

Abstract

Large language models frequently encounter conflicts between their parametric knowledge and contextual input, often resulting in factual inconsistencies or hallucinations. We propose Self-Reflective Debate for Contextual Reliability (SR-DCR), a lightweight framework that integrates token-level self-confidence with an asymmetric multi-agent debate to adjudicate such conflicts. A critic, deprived of context, challenges a defender who argues from the given passage; a judge model evaluates the debate and determines the context's reliability. The final answer is selected by combining the verdict with model confidence. Experiments on the ClashEval benchmark demonstrate that SR-DCR consistently enhances robustness to misleading context while maintaining accuracy on trustworthy inputs, outperforming both classical debate and confidence-only baselines with minimal computational overhead. The code is available at https://github.com/smiles724/Self-Reflective-Debates.