Holistic Evaluation and Failure Diagnosis of AI Agents

• URL: http://arxiv.org/abs/2605.14865v1
• Date: Thu, 14 May 2026 14:12:39 GMT
• Title: Holistic Evaluation and Failure Diagnosis of AI Agents
• Abstract summary: We present a holistic agent evaluation framework that pairs top-down diagnosis with bottom-up span-level evaluation.<n>This decomposition scales to traces of arbitrary length and produces span-level rationales for each verdict.<n>On the TRAIL benchmark, our framework achieves state-of-the-art results across all metrics.
• Score: 0.2922148761619506
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: AI agents execute complex multi-step processes, but current evaluation falls short: outcome metrics report success or failure without explaining why, and process-level approaches struggle to connect failure types to their precise locations within long, structured traces. We present a holistic agent evaluation framework that pairs top-down agent-level diagnosis with bottom-up span-level evaluation, decomposing analysis into independent per-span assessments. This decomposition scales to traces of arbitrary length and produces span-level rationales for each verdict. On the TRAIL benchmark, our framework achieves state-of-the-art results across all metrics on both GAIA and SWE-Bench, with relative gains over the strongest prior baselines of up to 38% on category F1, up to 3.5x on localization accuracy, and up to 12.5x on joint localization-categorization accuracy. Per-category analysis shows our framework leading in more error categories than any other evaluator. Notably, the same frontier model achieves several times higher localization accuracy when used inside our framework than as a monolithic judge over the full trace, showing that evaluation methodology, not model capability, is the bottleneck.

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)
• Beyond Binary Success: A Diagnostic Meta-Evaluation Framework for Fine-Grained Manipulation [98.79811866787263]
  We introduce MetaFine, a diagnostic meta-evaluation framework that disentangles manipulation competency along three axes.<n>We identify the visual encoder's ability to preserve local spatial structure as a key bottleneck for fine-grained precision.<n>By shifting evaluation from ranking to diagnosis, MetaFine turns benchmarking into an actionable compass for repairing the layered capacities underlying genuine physical dexterity.
  arXiv  Detail & Related papers  (2026-05-19T15:25:13Z)
• SAAG: Structured Agent Assessment and Grounding [13.62148061055744]
  We propose a cascaded diagnostic framework that decomposes agent-calling evaluation into three sequential stages.<n>We evaluate this framework on a benchmark derived from Glaive's function-calling dataset across registry sizes of 5, 10, and 15 agents.
  arXiv  Detail & Related papers  (2026-04-30T19:33:58Z)
• Claw-Eval: Toward Trustworthy Evaluation of Autonomous Agents [66.97968363332465]
  We introduce Claw-Eval, an end-to-end evaluation suite addressing all three gaps in agent benchmarks.<n>Claw-Eval comprises 300 human-verified tasks spanning 9 categories across three groups.<n>Every agent action is recorded through three independent evidence channels.
  arXiv  Detail & Related papers  (2026-04-07T17:43:18Z)
• RubricEval: A Rubric-Level Meta-Evaluation Benchmark for LLM Judges in Instruction Following [46.45323577110897]
  rubric-level evaluation has become a prevailing paradigm for evaluating instruction following in large language models (LLMs)<n>Our benchmark features: (1) the first rubric-level meta-evaluation benchmark for instruction following, (2) diverse instructions and responses spanning multiple instances and model sources, and (3) a substantial set of 3,486 quality-controlled subsets that better differentiates judge performance.<n>Considering evaluation paradigm, checklist-level evaluation outperforms rubric-level, explicit reasoning improves accuracy, and both together reduce inter-judge variance.
  arXiv  Detail & Related papers  (2026-03-26T07:55:32Z)
• From Comprehension to Reasoning: A Hierarchical Benchmark for Automated Financial Research Reporting [19.0993436440595]
  FinReasoning is a benchmark that decomposes Chinese research-report generation into three stages aligned with real analyst.<n>Based on the evaluation results, FinReasoning reveals that most models exhibit a understanding-execution gap.
  arXiv  Detail & Related papers  (2026-02-25T13:44:58Z)
• 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)
• Automated Skill Decomposition Meets Expert Ontologies: Bridging the Granularity Gap with LLMs [1.2891210250935148]
  This paper investigates automated skill decomposition using Large Language Models (LLMs)<n>Our framework standardizes the pipeline from prompting and generation to normalization and alignment with ontology nodes.<n>To evaluate outputs, we introduce two metrics: a F1-score that uses optimal embedding-based matching to assess content accuracy, and a hierarchy-aware F1-score that credits structurally correct placements to assess granularity.
  arXiv  Detail & Related papers  (2025-10-13T12:03:06Z)
• Continual Action Quality Assessment via Adaptive Manifold-Aligned Graph Regularization [53.82400605816587]
  Action Quality Assessment (AQA) quantifies human actions in videos, supporting applications in sports scoring, rehabilitation, and skill evaluation.<n>A major challenge lies in the non-stationary nature of quality distributions in real-world scenarios.<n>We introduce Continual AQA (CAQA), which equips with Continual Learning capabilities to handle evolving distributions.
  arXiv  Detail & Related papers  (2025-10-08T10:09:47Z)
• Where Did It All Go Wrong? A Hierarchical Look into Multi-Agent Error Attribution [0.7226144684379191]
  We present ECHO, a novel algorithm that combines hierarchical context representation, objective analysis-based evaluation, and consensus voting to improve error attribution accuracy.<n> Experimental results demonstrate that ECHO outperforms existing methods across various multi-agent interaction scenarios.
  arXiv  Detail & Related papers  (2025-10-06T15:07:13Z)
• SEOE: A Scalable and Reliable Semantic Evaluation Framework for Open Domain Event Detection [70.23196257213829]
  We propose a scalable and reliable Semantic-level Evaluation framework for Open domain Event detection.<n>Our proposed framework first constructs a scalable evaluation benchmark that currently includes 564 event types covering 7 major domains.<n>We then leverage large language models (LLMs) as automatic evaluation agents to compute a semantic F1-score, incorporating fine-grained definitions of semantically similar labels.
  arXiv  Detail & Related papers  (2025-03-05T09:37:05Z)
• GREAT Score: Global Robustness Evaluation of Adversarial Perturbation using Generative Models [60.48306899271866]
  We present a new framework, called GREAT Score, for global robustness evaluation of adversarial perturbation using generative models. We show high correlation and significantly reduced cost of GREAT Score when compared to the attack-based model ranking on RobustBench. GREAT Score can be used for remote auditing of privacy-sensitive black-box models.
  arXiv  Detail & Related papers  (2023-04-19T14:58:27Z)

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.