paper-with-me

Papers

Rethinking Atomic Decomposition for LLM Judges: A Prompt-Controlled Study of Reference-Grounded QA Evaluation

2026-03-30 · Xinran Zhang arxiv

Atomic decomposition -- breaking a candidate answer into claims before verifying each against a reference -- is a widely adopted design for LLM-based reference-grounded judges. However, atomic prompts are typically richer and longer, making it unclear whether any advantage comes from decomposition or from richer prompting. We study this for benchmark-style completeness-sensitive reference-support classification: classifying a candidate as fully supported, partially supported, or unsupported relative to a supplied reference. We compare a self-decomposing atomic judge (single-prompt decompose-and-verify) against a prompt-controlled holistic judge with the same inputs and a similarly detailed rubric. On 200 source examples per dataset across TruthfulQA, ASQA, and QAMPARI, with four model families, source-level paired tests, cluster bootstrap, and aggregation across three pre-frozen prompt variants per design family, we find the holistic judge matches or exceeds the atomic judge on two of three benchmarks: ASQA and QAMPARI favor holistic across all four families (statistically reliable in three of four), while TruthfulQA shows a small atomic edge. The holistic advantage is concentrated in partially\_supported cases -- incompleteness detection. A sensitivity check against human annotations confirms the ranking under both benchmark-completeness and human factual-correctness standards. Our finding is specific to the self-decomposing single-prompt pattern on three QA-style benchmarks with 200 source examples each; multi-stage atomic pipelines and non-QA tasks remain untested. Among perturbations examined, reference-quality degradation produced the largest accuracy drops for both judge families.

📄 PDF Abstract BibTeX arXiv:2603.28005

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Ask, Don't Judge: Binary Questions for Interpretable LLM Evaluation and Self-Improvement

2026-06-25 · Sangwoo Cho, Kushal Chawla, Pengshan Cai, Zefang Liu 외 arxiv

Evaluating LLM outputs remains a major bottleneck in NLP: human evaluation is expensive and slow, lexical metrics correlate poorly with human judgments on open-ended generation, and holistic LLM judges often produce opaq…

Prompt Attack Detection with LLM-as-a-Judge and Mixture-of-Models

2026-03-26 · Hieu Xuan Le, Benjamin Goh, Quy Anh Tang arxiv

Prompt attacks, including jailbreaks and prompt injections, pose a critical security risk to Large Language Model (LLM) systems. In production, guardrails must mitigate these attacks under strict low-latency constraints,…

Red Teaming

Do Encoders Suffice? A Systematic Comparison of Encoder and Decoder Safety Judges for LLM Adversarial Evaluation

2026-06-24 · Han Jeon, Shiv Medler, Joseph Voyles, Matt Wood arxiv

With the widespread adoption of large language models (LLMs) in chatbots and everyday applications, companies increasingly need guardrails that are effective while remaining low-cost and low-latency. Safety evaluation of…

MCJudgeBench: A Benchmark for Constraint-Level Judge Evaluation in Multi-Constraint Instruction Following

2026-05-05 · Jaeyun Lee, Junyoung Koh, Zeynel Tok, Hunar Batra 외 arxiv

Multi-constraint instruction following requires verifying whether a response satisfies multiple individual requirements, yet LLM judges are often assessed only through overall-response judgments. We introduce MCJudgeBenc…

Instruction Following

LLM Judges Have Dark Current: A Psychometric Datasheet for LLM-as-a-Judge Evaluation

2026-06-14 · Hiroyasu Usami, Keisuke Hara, Ayato Tsuboi, Naohiko Matsuda arxiv

LLM-as-a-judge systems are now routinely used for open-ended model evaluation, where human preference annotation is costly, slow, and difficult to reproduce. Yet these judges are often reported as scalar accuracy, win-ra…