Abstract
The transition from Large Language Models (LLMs) to agents shifts safety stakes from toxic text to irreversible environmental harm. While current defenses remain largely retrospective, proactive runtime intervention is bottlenecked by the lack of large-scale, causally-consistent data. We propose PROACT-Agent, a framework for synthesizing high-fidelity trajectories to enable real-time guardrails. We identify a critical "safety drift" in prior benchmarks, where lenient annotation paradigms fail to enforce temporal consistency. PROACT-Agent addresses this through: (1) Progressive Trajectory Unrolling to reveal risks hidden in long-context interactions; (2) Reasoning-Augmented Causal Rectification to enforce monotonic causal consistency; and (3) Culturally-Aware Data Localization for cross-border robustness. We introduce PROACT-Bench, a bilingual safety benchmark with 155,780 states labeled through multi-model adjudication. Evaluating updated context before the next LLM inference, the trained guard achieves 91.46% unsafe-class F1 and 90.63% exact-boundary detection under complete source holdout. In AgentDojo, it reduces non-DoS targeted attack success from 20.82% to 0.40%.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Jia, D., Liu, W., Du, X., Li, Y., Yang, Y., Yu, H., Zhan, Z., & Zhou, C. (2026). PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety. https://omanscience.com/en/articles/proact-agent-progressive-runtime-oversight-and-active-circuit-breaking-for-real-time-safety
MLA 9
Jia, Ding, et al. "PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety." https://omanscience.com/en/articles/proact-agent-progressive-runtime-oversight-and-active-circuit-breaking-for-real-time-safety.
Chicago (author–date)
Jia, Ding, Wei Liu, Xianglong Du, Yingjie Li, Yingqing Yang, Huili Yu, Zhangsong Zhan, and Chu Zhou. 2026. "PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety." https://omanscience.com/en/articles/proact-agent-progressive-runtime-oversight-and-active-circuit-breaking-for-real-time-safety.
Harvard
Jia, D., Liu, W., Du, X., Li, Y., Yang, Y., Yu, H., Zhan, Z. and Zhou, C. (2026) 'PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety', Available at: https://omanscience.com/en/articles/proact-agent-progressive-runtime-oversight-and-active-circuit-breaking-for-real-time-safety.
Vancouver
Jia D, Liu W, Du X, Li Y, Yang Y, Yu H, et al. PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety. https://omanscience.com/en/articles/proact-agent-progressive-runtime-oversight-and-active-circuit-breaking-for-real-time-safety
IEEE
D. Jia, W. Liu, X. Du, Y. Li, Y. Yang, H. Yu, Z. Zhan, and C. Zhou, "PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety," https://omanscience.com/en/articles/proact-agent-progressive-runtime-oversight-and-active-circuit-breaking-for-real-time-safety.