The Double Measurement Confound in Agent Benchmarks: De-Scaffolding, Ground-Truth Scoring, and Reliability Beyond the Mean

• URL: http://arxiv.org/abs/2609.09218v1
• Date: Sun, 06 Sep 2026 21:23:11 GMT
• Title: The Double Measurement Confound in Agent Benchmarks: De-Scaffolding, Ground-Truth Scoring, and Reliability Beyond the Mean
• Abstract summary: Agent benchmarks are increasingly used to compare large language models (LLMs) and guide deployment decisions.<n>We identify a double measurement confound: execution-critical decisions are performed by a fixed scaffold instead of the model, while the scorer evaluates outputs using criteria that may not reflect task correctness.<n>We unify these issues within a measurement-theoretic framework that characterizes when benchmark scores can be interpreted as evidence of model capability.
• Score: 2.822987192839875
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Agent benchmarks are increasingly used to compare large language models (LLMs) and guide deployment decisions, yet benchmark scores are meaningful only if they measure model capability rather than properties of the evaluation pipeline. We identify a double measurement confound: execution-critical decisions are performed by a fixed scaffold instead of the model, while the scorer evaluates outputs using criteria that may not reflect task correctness. We unify these issues within a measurement-theoretic framework that characterizes when benchmark scores can be interpreted as evidence of model capability, and instantiate it with an audit-and-repair protocol that (i) transfers execution-critical decisions from the scaffold to the model, (ii) replaces shape-based evaluation with seeded ground-truth scoring, and (iii) reports reliability beyond the mean through worst-case and tail-risk metrics. Experiments on ComtradeBench show that the joint intervention transforms a nearly flat leaderboard into a reliability spectrum that distinguishes both average performance and robustness across seeds. Applying the audit to existing benchmarks further shows that scorer validity is benchmark-specific, whereas scaffold ownership is an uncontrolled axis wherever we probed it. Our results suggest that benchmark scores should be interpreted together with their scaffolding level, scoring criterion, and reliability profile, providing a practical framework for more valid evaluation of LLM agents.

Related papers

• Position: Evaluation Scores Are Perishable Knowledge Claims [0.0]
  We argue that evaluation scores should be treated as claims with three properties: formality, scope, and validity windows.<n>We illustrate the cost of mean aggregation on the public HELM leaderboard.
  arXiv  Detail & Related papers  (2026-07-28T18:50:39Z)
• DualEval: Joint Model-Item Calibration for Unified LLM Evaluation [67.99119926012686]
  DualEval is a latent model-item calibration framework that represents models and evaluation items in a shared space.<n>We apply DualEval across four domains: coding, math, miscellaneous domain-knowledge tasks, and generic everyday user queries.
  arXiv  Detail & Related papers  (2026-06-24T22:40:46Z)
• Latent Performance Profiling of Large Language Models [47.009623327601226]
  We introduce Latent Performance Profiling (LPP) -- a framework that derives task-agnostic diagnostics from hidden activations and output distributions.<n>Unlike static accuracy scores, LPP provides stable, architecture-sensitive signatures across models of similar size.<n>We show that models with similar benchmark scores can exhibit contrasting latent profiles, such as differences in entropy or adaptability.
  arXiv  Detail & Related papers  (2026-05-28T14:41:26Z)
• Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents? [61.49434544687523]
  We introduce REFLECT, a meta-evaluation benchmark targeting fine-grained failure detection in agentic environments.<n> REFLECT defines a detailed taxonomy of process- and outcome-level failure modes, instantiated by performing controlled and localized interventions.<n>Our experiments show that even the best-performing models achieve overall accuracies below 55% across reasoning, tool-use, and report-quality failures.
  arXiv  Detail & Related papers  (2026-05-18T23:55:08Z)
• Navigating the Sea of LLM Evaluation: Investigating Bias in Toxicity Benchmarks [0.36748639131154315]
  This work investigates the robustness of established benchmarking setups and examines how to measure intrinsic biases.<n>Our experiments uncover significant discrepancies in benchmark behaviors when evaluation setups are altered.<n>We observe model-specific instabilities, demonstrating a clear need for more robust and comprehensive safety evaluation frameworks.
  arXiv  Detail & Related papers  (2026-05-11T14:27:39Z)
• Deployment-Relevant Alignment Cannot Be Inferred from Model-Level Evaluation Alone [11.663456969895462]
  Alignment evaluation in machine learning has largely become evaluation of models.<n>This paper argues that deployment-relevant alignment cannot be inferred from model-level evaluation alone.
  arXiv  Detail & Related papers  (2026-05-06T03:28:30Z)
• 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)
• Benchmark^2: Systematic Evaluation of LLM Benchmarks [66.2731798872668]
  We propose Benchmark2, a comprehensive framework comprising three complementary metrics.<n>We conduct experiments across 15 benchmarks spanning mathematics, reasoning, and knowledge domains.<n>Our analysis reveals significant quality variations among existing benchmarks and demonstrates that selective benchmark construction can achieve comparable evaluation performance.
  arXiv  Detail & Related papers  (2026-01-07T14:59:03Z)
• The Flaw of Averages: Quantifying Uniformity of Performance on Benchmarks [32.00464870277127]
  We study benchmark reliability from a distributional perspective and introduce benchmark harmony.<n>High harmony is a desirable benchmark property, indicating that the aggregate metric reflects uniform competence across models.<n>By recommending that harmony should be reported alongside accuracy, we reframe evaluation from simple performance averages to a more robust, distributionally reliable measurement of performance.
  arXiv  Detail & Related papers  (2025-09-30T02:14:30Z)
• 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)
• Generating Benchmarks for Factuality Evaluation of Language Models [61.69950787311278]
  We propose FACTOR: Factual Assessment via Corpus TransfORmation, a scalable approach for evaluating LM factuality. FACTOR automatically transforms a factual corpus of interest into a benchmark evaluating an LM's propensity to generate true facts from the corpus vs. similar but incorrect statements. We show that: (i) our benchmark scores increase with model size and improve when the LM is augmented with retrieval; (ii) benchmark score and perplexity do not always agree on model ranking; (iii) when perplexity and benchmark score disagree, the latter better reflects factuality in open-ended generation.
  arXiv  Detail & Related papers  (2023-07-13T17:14:38Z)

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.