cloud-iam-least-privilege-bench / baseline_model_eval.csv
soham27's picture
Release Cloud IAM Wildcard & Privilege Escalation Bench v1.0
00c47d6 verified
Raw History Blame Contribute Delete
326 Bytes
model_architecture,domain_accuracy,p95_latency_ms,citation_precision,hallucination_rate_pct
Botmartz Deterministic State Pipeline,0.964,145,0.985,0.2%
LlamaParse AST + Hybrid RRF (k=60),0.928,380,0.942,1.4%
Naive Dense Vector Embeddings (Cosine),0.582,190,0.612,18.6%
PyPDF Heuristic OCR + LangChain,0.415,85,0.435,34.8%