janus-foreseeing-latent-risk-for-long-horizon-agen

name	janus-foreseeing-latent-risk-for-long-horizon-agen
description	Skill generated from arXiv paper 2607.19913: JANUS: Foreseeing Latent Risk for Long-Horizon Agent Safety
metadata	{"arxiv":{"id":"2607.19913","title":"JANUS: Foreseeing Latent Risk for Long-Horizon Agent Safety","authors":["Yuan Xiong","Linji Hao","Shizhu He","Yequan Wang","Lijun Li"],"published":"2026-07-22","categories":["cs.AI","cs.CL","cs.CR"],"url":"https://arxiv.org/abs/2607.19913","utility":1}}

JANUS: Foreseeing Latent Risk for Long-Horizon Agent Safety

arXiv: 2607.19913
Published: 2026-07-22
Authors: Yuan Xiong, Linji Hao, Shizhu He, Yequan Wang, Lijun Li
Categories: cs.AI, cs.CL, cs.CR
Utility: 1.00

Key Innovation

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 fro...

Potential Application

This paper presents advancements that could be applied to enhance agent capabilities in the areas of cs.AI, cs.CL, cs.CR.

References

• arXiv: https://arxiv.org/abs/2607.19913