Yunhao-Feng/AgentHazard
AgentHazard A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents ๐ Website | ๐ Dataset | ๐ Paper | ๐ Appendix ๐ฏ Overview AgentHazard is a comprehensive benchmark for evaluating harmful behavior in computer-use agents. Unlike traditional prompt-level safety benchmarks, AgentHazard focuses on execution-level failures that emerge through the composition of locally plausible steps across multi-turn, tool-mediated trajectories.โฆ See the full description on the dataset page: https://huggingface.co/datasets/Yunhao-Feng/AgentHazard.
| Author / organization | Yunhao-Feng |
|---|---|
| Downloads | 238 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 7 April 2026 |
| Last updated at source | 7 April 2026 |
Popularity, last 3 days
238 now (-3 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
Recent changes
- 16 Oct 2026 Description: AgentHazard A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents ๐ Website | ๐ Dataset | ๐ Paper | ๐ Appendix ๐ฏ Overview AgentHazard is a comprehensive benchmark for evaluating harmful behavior in computer-use agents. Unl โ AgentHazard A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents ๐ Website | ๐ Dataset | ๐ Paper | ๐ Appendix ๐ฏ Overview AgentHazard is a comprehensive benchmark for evaluating harmful behavior in computer-use agents. Unl
Changes detected by Silytics between refreshes of the source.
Related papers
More from Yunhao-Feng
- Yunhao-Feng/AdaptiveSafetyDataset
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.