Predictive Safety Guardrails Do Not Always Make LLM Agents Safer
cs.AI
Submitted: 2026-11-02
Updated: 2026-11-02
Code: https://github.com/REH502/rethinking-predictive-safety
Terminology
Sources
- GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
- Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation
- SafePred: A Predictive Guardrail for Computer-Using Agents via World Models
- General Agentic Planning Through Simulative Reasoning with World Models
- Is Your LLM Secretly a World Model of the Internet? Model-Based Planning for Web Agents
- World Models
- Mastering Diverse Domains through World Models
- From Word to World: Can Large Language Models be Implicit Text-based World Models?
- DreamGuard: Efficient Runtime Guardrail for LLM Agents via Risk-Aware World Model
- Auditing Agent Harness Safety
- IS-Bench: Evaluating Interactive Safety of VLM-Driven Embodied Agents in Daily Household Tasks
- Current Agents Fail to Leverage World Model as Tool for Foresight
- Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting
- ProbGuard: Proactive Runtime Monitoring for LLM Agent Safety via Probabilistic Prediction
- SafeMCP: Proactive Power Regulation for LLM Agent Defense via Environment-Grounded Look-Ahead Reasoning
- OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
- JANUS: Foreseeing Latent Risk for Long-Horizon Agent Safety
- SeerGuard: A Safety Framework for Mobile GUI Agents via World Model Prediction
- Agent-SafetyBench: Evaluating the Safety of LLM Agents
- WebArena: A Realistic Web Environment for Building Autonomous Agents
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents