Raucle-bench leaderboard
165 prompts across benign=40, data_exfiltration=20, direct_injection=25, evasion=20, indirect_injection=15, jailbreak=25, tool_abuse=20. Run 2026-05-14-v0.1-baseline.
| Adapter | F1 | Detection rate | FPR | Strict match | p50 latency | p99 latency |
|---|---|---|---|---|---|---|
always-block-baseline v1.0.0 | 0.862 | 100.0% | 100.0% | 73.3% | 0.00 ms | 0.00 ms |
raucle-detect-strict v0.5.0 | 0.523 | 36.0% | 5.0% | 37.0% | 0.03 ms | 0.05 ms |
raucle-detect-standard v0.5.0 | 0.264 | 15.2% | 0.0% | 24.9% | 0.03 ms | 0.05 ms |
raucle-detect-permissive v0.5.0 | 0.016 | 0.8% | 0.0% | 24.2% | 0.03 ms | 0.07 ms |
always-allow-baseline v1.0.0 | 0.000 | 0.0% | 0.0% | 24.2% | 0.00 ms | 0.00 ms |