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

• URL: http://arxiv.org/abs/2603.28005v1
• Date: Mon, 30 Mar 2026 03:55:26 GMT
• Title: Rethinking Atomic Decomposition for LLM Judges: A Prompt-Controlled Study of Reference-Grounded QA Evaluation
• Abstract summary: 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.<n>Our finding is specific to the self-decomposing single-prompt pattern on three QA-style benchmarks with 200 source examples each.
• Score: 5.94231111588812
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• AtomCite: Verification and Correction of Supplied Page-Level Citations in Multi-Page Documents [21.842813905857863]
  AtomCite parses an answer into claims, checks each claim against the image of the cited page, and applies a deterministic repair policy.<n>It reaches around 93% binary verification accuracy on the injected benchmark, significantly outperforming every OCR-only condition.<n>Its repair policy lifts citation precision on the injected mix from a constructed 34% to 87-90% while retaining over 90% of correct claims.
  arXiv  Detail & Related papers  (2026-09-05T01:31:50Z)
• What Does Your Short-Answer VQA Score Actually Measure? Evaluator-Dependent Instability in Multimodal Short-Answer Benchmarks [25.519850997497283]
  Short-answer VQA benchmarks conflate two distinct quantities.<n>We use a human-validated semantic judge to audit over 37k official errors.
  arXiv  Detail & Related papers  (2026-07-11T10:01:29Z)
• CriterAlign: Criterion-Centric Rationale Alignment for Code Preference Judging [95.02210956333374]
  We propose a criterion-centric framework that adapts rubric-based judging to pairwise preference evaluation.<n>On BigCodeReward, CriterAlign improves a Qwen2.5-VL-32B monolithic judge from 60.4% to 66.3% accuracy.
  arXiv  Detail & Related papers  (2026-05-19T10:59:19Z)
• C2-Faith: Benchmarking LLM Judges for Causal and Coverage Faithfulness in Chain-of-Thought Reasoning [0.6138671548064355]
  Large language models (LLMs) are increasingly used as judges of chain-of-thought (CoT) reasoning.<n>We introduce C2-Faith, a benchmark that targets two complementary dimensions of faithfulness: causality (does each step logically follow from prior context?) and coverage.<n>We evaluate three frontier judges under three tasks: binary causal detection, causal step localization, and coverage scoring.
  arXiv  Detail & Related papers  (2026-03-05T13:36:47Z)
• IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation [85.56193980646981]
  We propose IF-RewardBench, a comprehensive meta-evaluation benchmark for instruction-following.<n>For each instruction, we construct a preference graph containing all pairwise preferences among multiple responses.<n>Experiments on IF-RewardBench reveal significant deficiencies in current judge models.
  arXiv  Detail & Related papers  (2026-03-05T02:21:17Z)
• Autorubric: A Unified Framework for Rubric-Based LLM Evaluation [34.429649156970015]
  We present a unified framework for evaluating large language models (LLMs)<n>Each identified technique is paired with its realization in Autorubric, an open-source Python framework proposed in this paper.<n>Autorubric supports binary, ordinal, and nominal criteria with weights; single-judge and multi-judge ensemble evaluation with majority, weighted, unanimous, and any-vote aggregation.
  arXiv  Detail & Related papers  (2026-02-13T02:26:30Z)
• Reference-Free Rating of LLM Responses via Latent Information [53.463883683503106]
  We study the common practice of asking a judge model to assign Likert-scale scores to free-text responses.<n>We then propose and evaluate Latent Judges, which derive scalar ratings from internal model signals.<n>Across a broad suite of pairwise and single-rating benchmarks, latent methods match or surpass standard prompting.
  arXiv  Detail & Related papers  (2025-09-29T12:15:52Z)
• Pretraining on the Test Set Is No Longer All You Need: A Debate-Driven Approach to QA Benchmarks [2.3188831772813105]
  We propose a debate-driven evaluation paradigm that transforms any existing QA dataset into structured adversarial debates.<n>We make two main contributions: (1) an evaluation pipeline to systematically convert QA tasks into debate-based assessments, and (2) a public benchmark that demonstrates our paradigm's effectiveness on a subset of MMLU-Pro questions.
  arXiv  Detail & Related papers  (2025-07-23T17:58:14Z)
• JudgeRank: Leveraging Large Language Models for Reasoning-Intensive Reranking [81.88787401178378]
  We introduce JudgeRank, a novel agentic reranker that emulates human cognitive processes when assessing document relevance. We evaluate JudgeRank on the reasoning-intensive BRIGHT benchmark, demonstrating substantial performance improvements over first-stage retrieval methods. In addition, JudgeRank performs on par with fine-tuned state-of-the-art rerankers on the popular BEIR benchmark, validating its zero-shot generalization capability.
  arXiv  Detail & Related papers  (2024-10-31T18:43:12Z)
• Localizing Factual Inconsistencies in Attributable Text Generation [74.11403803488643]
  We introduce QASemConsistency, a new formalism for localizing factual inconsistencies in attributable text generation.<n>We show that QASemConsistency yields factual consistency scores that correlate well with human judgments.
  arXiv  Detail & Related papers  (2024-10-09T22:53:48Z)
• Large Language Models are not Fair Evaluators [60.27164804083752]
  We find that the quality ranking of candidate responses can be easily hacked by altering their order of appearance in the context. This manipulation allows us to skew the evaluation result, making one model appear considerably superior to the other. We propose a framework with three simple yet effective strategies to mitigate this issue.
  arXiv  Detail & Related papers  (2023-05-29T07:41:03Z)

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.