Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?

Leyao Wang, Yanan He, Peng Chen, Asaf Yehudai, Yixin Liu, Rex Ying, Michal Shmueli-Scheuer, Arman Cohan

arXiv:2605.19196 · 2026-05-20 공개 · arXiv · PDF

llm-evaluation tool-use llm-as-judge meta-evaluation reasoning-quality research-agents agent-execution evidence-verification

Abstract

Deep research agents increasingly automate complex information-seeking tasks, producing evidence-grounded reports via multi-step reasoning, tool use, and synthesis. Their growing role demands scalable, reliable evaluation, positioning LLM-as-judge as a supervision paradigm for assessing factual accuracy, evidence use, and reasoning quality. Yet the reliability of these judges for deep research agents remains poorly understood, posing a critical meta-evaluation problem: before deploying LLM judges to supervise research agents, we must first evaluate the judges themselves. Existing meta-evaluations fall short in two ways: (1) reliance on coarse, subjective human-preference agreement; (2) focus on instruction-following or verifiable tasks, leaving open-ended agent executions unexplored. To address these gaps, we introduce REFLECT (REliable Fine-grained LLM judge Evaluation via Controlled inTervention), a meta-evaluation benchmark targeting fine-grained failure detection in agentic environments. REFLECT defines a detailed taxonomy of process- and outcome-level failure modes, instantiated by performing controlled and localized interventions on quality-screened agent execution traces. This yields verifiable, comprehensive, and fine-grained instances for validating the judge models. Our experiments show that current LLM judges remain unreliable: even the best-performing models achieve overall accuracies below 55% across reasoning, tool-use, and report-quality failures, with especially poor performance on evidence verification. Together, our taxonomy and findings expose systematic judge limitations, reveal tradeoffs in cost and reliability, and offer actionable guidance for building more reliable evaluation pipelines for deep research agents.

한국어 요약

📋 한 줄 요약

**[LLM-as-judge 메타평가 / Deep Research Agent]** 통제된 개입으로 만든 fine-grained 실패 사례 위에서 LLM 판정자의 신뢰성을 진단하는 메타평가 벤치마크 REFLECT 제안.

🎯 핵심 기여도

💡 핵심 아이디어

LLM 판정자의 신뢰성을 평가하려면 사람 선호도 일치 같은 거친 신호가 아니라, "어떤 종류의 실패가 어디서 발생했는지"를 통제된 개입으로 만들어 정답이 명확한 사례 위에서 측정해야 한다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: deep research agent 평가 파이프라인 구축에 실증적·실행 가능한 가이드 제공, LLM-as-judge 신뢰성을 정량적으로 다루는 메타평가 표준 제시. **한계**: 통제된 개입이 자연 발생 실패 분포를 완전히 대표하지 못할 수 있으며, taxonomy의 범위는 에이전트 아키텍처 발전에 따라 갱신 필요.

🚀 실용적 활용