CORTEXA
← Browse
arxivcs.AIcs.CLcs.CR2026-07-22

JANUS: Foreseeing Latent Risk for Long-Horizon Agent Safety

Yuan Xiong, Linji Hao, Shizhu He, Yequan Wang, Lijun Li

Agent safety is moving from content moderation toward preventing operational failures before tool-using agents act. We propose Janus, a foresight-oriented framework for long-horizon agent safety that trains guards to anticipate delayed risks from partial trajectories. Janus synthesizes diverse agent trajectories via multi-agent simulation and learns a shared policy with two coupled tasks: an anticipation task that forecasts safety-relevant futures and an adjudication task that decides safety from both the observed prefix and anticipated future. The two tasks are jointly optimized with CoAA-RL, which rewards forecasts by their utility for downstream safety judgment. The resulting guard model, Vanguard, blocks unsafe actions before execution. Across four agent-safety benchmarks, Vanguard improves average protection by 15.9 percentage points over baseline guards while increasing benign task completion by 5.1 percentage points.

View free PDFSource page

Related papers

arxivcs.LGcs.AIcs.CLcs.CR2026-07-08

Efficient Safety Alignment of Language Models via Latent Personality Traits

Mohamed Amine Merzouk, Nolan Smyth, Damiano Fornasiere, Linh Le, David Williams-King, Adam Oberman

Current safety methods for large language models are known to be vulnerable to adversarial attacks, motivating research into robust alternatives. Latent Adversarial Training (LAT) is among the most effective defenses, but can degrade utility and requires training on large dataset…

View free PDFSource page
arxivcs.AIcs.CLcs.CR2026-07-08

The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving Agents

Xing Zhang, Yanwei Cui, Guanghui Wang, Ziyuan Li, Wei Qiu, Bing Zhu, et al.

A self-evolving agent retires its bad skills by watching them fail, so what happens when the judge cannot see the failures? Skill retirement is the structural constraint that keeps a growing library from drifting below the no-skill baseline, but its guarantee assumes an unbiased…

View free PDFSource page
arxivcs.CRcs.AIcs.CL2026-07-20

JailMeter: An Evidence-Based Evaluation Framework for Jailbreak Attacks on Large Language Models

Qingjia Huang, Jingyu Zhang, Jianguo Wu, Yakai Li, Weijuan Zhang, Yankai Rong, et al.

The assessment of jailbreak attacks against large language models currently suffers from inconsistent evaluation criteria and methods, leading to unreliable estimates of attack success rates. We propose JailMeter, an evidence-based evaluation framework designed to more faithfully…

View free PDFSource page
arxivcs.CLcs.AIcs.CR2026-07-21

Find Before You Fine-Tune: A Diagnostic Study of Small LLMs for Cybersecurity QA

Shaswata Mitra, Subash Neupane, Trisha Chakraborty, Himanshu Tripathi, Sudip Mittal, Aritran Piplai, et al.

Large Language Models (LLMs) are increasingly fine-tuned for critical-domain Question-Answering (QA), yet choosing which small model to adapt, before paying the cost of adaptation, remains difficult. Fine-tuning can improve domain alignment, but it may also erode prior knowledge,…

View free PDFSource page
arxivcs.CRcs.AIcs.CLcs.MA2026-07-20

Self-State Attacks on Self-Hosted AI Agents: How Far Can OS Defenses Go?

Yimeng Chen, Nathanaël Denis, Roberto Di Pietro, Jürgen Schmidhuber

Self-hosted AI agents read and write their own memory and configuration files to function. An agent may get compromised via corruption of its own state -- a compromise realized via legitimate OS system call invocation. We refer to this class of threats as self-state attacks. In t…

View free PDFSource page