Metraipolis › archive › dc_f5c8a4bb0c32
What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation — arXiv:2610.30491
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.30491
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-23T15:12:52.837Z
- SHA-256 of the archived text
8f2d8806436ff5e2b89b6954be78d070a545fe1a8786fa28ed890fb9fd485ef3- Stable URL of this page
https://metraipolis.com/documento/dc_f5c8a4bb0c32
How to cite
What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation — arXiv:2610.30491 — arXiv (en), https://arxiv.org/abs/2610.30491 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f5c8a4bb0c32 (Metraipolis archive dc_f5c8a4bb0c32, sha256 8f2d8806436ff5e2b89b6954be78d070a545fe1a8786fa28ed890fb9fd485ef3)
BibTeX
@misc{dcf5c8a4bb0c32,
title = {What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation — arXiv:2610.30491},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.30491},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f5c8a4bb0c32; sha256 8f2d8806436ff5e2b89b6954be78d070a545fe1a8786fa28ed890fb9fd485ef3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.30491. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM agents are often evaluated by reducing execution traces to labels such as final-answer correctness, tool verification, and fault propagation. These labels are measurements produced by parsers, behavioral detectors, and harness rules, and they may not fully capture what the trace shows. We study this in a tool-use setting built on GSM8K, where agents call a calculator whose outputs are occasionally corrupted, giving clear ground truth for what the tool returned, what the model later used, and what the detectors reported. We audit three trace-level detectors across six production models and 30 problems per model. Final correctness, visible re-verification, and fault propagation can disagree in systematic ways. A key failure mode is silent tool-output override: the calculator returns a corrupted value, but the model later uses a different, often correct, value without calling the tool again, so the bad value stops propagating while a verification detector records no re-check. Human review of 90 traces agrees perfectly with the automated labels for final correctness (67/67), fault impact (19/19), and fault relevance (18/18), but only 67/81 for verification; all 14 disagreements fall into four recurring patterns. Execution limits also change the reported result: in an ablation across two models, one ranges from 21/30 to 27/30 correct as step and token budgets change, while the other stays between 27/30 and 29/30. These findings motivate measurement transparency: an automated label should come with the trace evidence behind it, whether the question the detector asks applied to that trace, whether the run was cut off by an execution limit, and any human corrections. Our trace viewer makes this information inspectable alongside the reported results. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI) ; Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2610.30491 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-23T15:09:08.367Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).