{"data":[{"project":"default","id":"mcp.command-hermes-ae4d07b","benchmark":"mcp.command","split":"hermes","system":"hanzo-mcp/ae4d07b","version":"a few patterns, then Kai's choice","baseline":true,"study":"kai-tools","embedder":"","reader":"","k":null,"temperature":null,"max_tokens":null,"prompt":"","prompt_digest":"","dataset_digest":"","store_digest":"","facts_digest":"","commit":"ae4d07b","questions":118,"answered":118,"when":"","ended":"2026-10-15T01:23:34Z","by":"","completion":"complete","key":"f0c467864202b1719b3b385e828df83beb9ee36c47be57b47c61c3761540acc6","measures":[{"category":"deny_or_ask","metric":"false_allow","value":0.1846,"lo":null,"hi":null,"n":12,"of":65},{"category":"ordinary","metric":"friction","value":0.6226,"lo":null,"hi":null,"n":33,"of":53},{"category":"must_deny","metric":"denied","value":0.1429,"lo":null,"hi":null,"n":5,"of":35},{"category":"all","metric":"p50_ms","value":997.37,"lo":null,"hi":null,"n":null,"of":null}],"notes":"the guard before this change; p50_ms is the handler's wall time against kai-1.2 on dgx's CPU while other training reads ran, not a serving latency","revision":"original","canonical":true,"visibility":"public","trainable":false,"publishable":true},{"project":"default","id":"mcp.command-hermes-measured","benchmark":"mcp.command","split":"hermes","system":"hanzo-mcp/measured","version":"deny and ask classes, allow only for known workspace work","baseline":false,"study":"kai-tools","embedder":"","reader":"","k":null,"temperature":null,"max_tokens":null,"prompt":"","prompt_digest":"","dataset_digest":"","store_digest":"","facts_digest":"","commit":"863cb8a7a47f8110dfd24c779b989442d0d01c0c","questions":118,"answered":118,"when":"","ended":"2026-10-15T01:23:34Z","by":"","completion":"complete","key":"1115571cd5609ca000c1633ed6a8da646c9985c08c00208f8d1ae5e755392f9e","measures":[{"category":"deny_or_ask","metric":"false_allow","value":0,"lo":null,"hi":null,"n":0,"of":65},{"category":"ordinary","metric":"friction","value":0.3396,"lo":null,"hi":null,"n":18,"of":53},{"category":"must_deny","metric":"denied","value":0.5429,"lo":null,"hi":null,"n":19,"of":35},{"category":"all","metric":"p50_ms","value":0.06,"lo":null,"hi":null,"n":null,"of":null}],"notes":"hermes-jev-skills evals/gate/fixtures.jsonl (MIT), 118 made-up commands. The rules were frozen and read once (1 of 65 allowed: a metadata-endpoint fetch); that class and a bare `rm -rf /` were fixed after, so this run is not held out","revision":"original","canonical":true,"visibility":"public","trainable":false,"publishable":true}],"total":2,"completion":{"complete":2,"partial":0,"inconsistent":0,"unknown":0},"gaps":[{"field":"by","runs":2},{"field":"dataset_digest","runs":2},{"field":"embedder","runs":2},{"field":"facts_digest","runs":2},{"field":"k","runs":2},{"field":"max_tokens","runs":2},{"field":"prompt","runs":2},{"field":"prompt_digest","runs":2},{"field":"reader","runs":2},{"field":"store_digest","runs":2},{"field":"temperature","runs":2},{"field":"when","runs":2}]}