MLZoo/CrossMCP-Bench
CrossMCP-Bench Authorization-conditioned benchmark for evaluating runtime security guards for tool-calling AI agents over multi-server MCP architectures. Dataset Description CrossMCP-Bench comprises 134 core scenarios (75 attacks, 59 benign) spanning 6 policy categories, plus supplementary scenarios: 70 coverage-expansion, 5 whitebox, 10 LLM-adversarial, and 53 independent benign for FPR validation. Each scenario includes an authorization context that determines… See the full description on the dataset page: https://huggingface.co/datasets/MLZoo/CrossMCP-Bench.
| Author / organization | MLZoo |
|---|---|
| License (as declared) | cc-by-4.0 |
| Downloads | 174 (as reported by Hugging Face) |
| Likes | 1 |
| Created | 3 May 2026 |
| Last updated at source | 3 May 2026 |
Popularity, last 3 days
174 now. Counts as reported by the source, snapshotted daily by Silytics.
Recent changes
- 16 Oct 2026 Description: CrossMCP-Bench Authorization-conditioned benchmark for evaluating runtime security guards for tool-calling AI agents over multi-server MCP architectures. Dataset Description CrossMCP-Bench comprises 134 core scenarios (75 attacks, 59 benign → CrossMCP-Bench Authorization-conditioned benchmark for evaluating runtime security guards for tool-calling AI agents over multi-server MCP architectures. Dataset Description CrossMCP-Bench comprises 134 core scenarios (75 attacks, 59 benign
Changes detected by Silytics between refreshes of the source.
More from MLZoo
- MLZoo/edu-fineweb-10BDataset
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 Information may be incomplete or out of date.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.