HazardAuditor: From Executable Threats to Safer Computer-Use Agents
HazardAuditor trains execution-grounded guard models for computer-use agents, improving safety verdict accuracy by up to 16.5 points.
HazardAuditor runs heterogeneous agents (Claude Code, Codex, Hermes, OpenClaw) in controlled environments and normalizes their interactions into a canonical event representation for cross-framework supervision. It introduces Guard Policy Optimization (GuardPO), which converts deterministic safety outcomes into sequence-level advantages and normalizes rationale and verdict regions so the safety decision becomes the effective optimization unit. Across multiple benchmarks and heterogeneous computer-use systems, HazardAuditor improves accuracy by up to 16.5 percentage points over the strongest prior guard model. Code, models, and evaluation artifacts are being released.
- Execution-grounded supervision for computer-use agent safety, not just prompts
- Normalizes heterogeneous agent traces into canonical event representation
- GuardPO fixes token-level objective mismatch in generative guard training
- Up to 16.5 percentage points accuracy gain over prior guard models
From the original article
1 min read at huggingface.co
Computer-use agents increasingly interact with browsers, terminals, file systems, and external services, introducing safety risks that emerge through runtime behavior rather than generated content alone.
By Yunhao Feng, Ruixiao Lin, Ming Wen · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.