EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

TraceWeave: Policy-Conditioned Assessment of Computer-Use Agents

Yunhao Feng, Yuxiang Xie, Xiaolong Li, Yanming Guo

Electronics · 17.11.2026 · 15 (22) · 5351

DOI: 10.3390/electronics15225351 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Electronics, online erschienen am 17.11.2026. Verfasst von Yunhao Feng und 3 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Fund of Laboratory for Big Data and Decision.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Computer-use agents (CUAs) are increasingly used to automate tasks across digital environments, raising concerns about the safety of their actions and consequences. Safety requirements differ across deployment settings, where the same behavior may be permitted under one policy and prohibited under another. This variation calls for safety assessors that evaluate agent behavior against application-specific policies rather than rely solely on predefined harm categories. We introduce TraceWeave, an integrated framework for constructing policy-conditioned safety data and training corresponding guard models. We release AdaptiveSafety, a dataset of 11,939 examples pairing user-defined policies with interaction records, supporting explanations, and complete sets of violated rules. Its construction combines interactions from multiple sources with controlled policy and behavioral variations, followed by evidence-grounded annotation, verification, and adjudication. Using AdaptiveSafety, we train generative guard models through supervised fine-tuning to assess policy compliance and identify violations with supporting explanations. Evaluation covers three model scales on AdaptiveSafety and an additional policy-conditioned benchmark. On the AdaptiveSafety validation split, the 4-B model achieves 89.90% accuracy in assessing policy compliance and correctly predicts the complete set of violated rules in 77.20% of cases. Further analysis shows that correct compliance judgments can still be accompanied by incomplete or incorrect identification of the violated rules. These findings highlight a limitation of aggregate safety classification and establish a reference for developing more precise policy-conditioned assessors. TraceWeave and AdaptiveSafety support research on how safety models interpret interaction evidence and apply changing policies.

Förderung

  • Fund of Laboratory for Big Data and Decision

Zitieren

APA

Feng, Y., Xie, Y., Li, X., & Guo, Y. (2026). TraceWeave: Policy-Conditioned Assessment of Computer-Use Agents. Electronics, 15(22), 5351. https://doi.org/10.3390/electronics15225351

Vancouver

Feng Y, Xie Y, Li X, Guo Y. TraceWeave: Policy-Conditioned Assessment of Computer-Use Agents. Electronics. 2026;15(22):5351. doi:10.3390/electronics15225351

BibTeX
@article{Feng2026,
  author = {Feng, Yunhao and Xie, Yuxiang and Li, Xiaolong and Guo, Yanming},
  title = {TraceWeave: Policy-Conditioned Assessment of Computer-Use Agents},
  journal = {Electronics},
  year = {2026},
  volume = {15},
  number = {22},
  pages = {5351},
  doi = {10.3390/electronics15225351}
}
Kostenlos starten →