ScamSentry / eval

Commit History

  1. Day 6: fix batch2 labels, fix LLM-judge verdict bug, add pass_k and rpd diagnostic scripts

    SooryaKiranKakkarayil
    6b6557c

  1. Day 4 complete: 3-layer injection defense, 15/15 on adversarial suite, transient-error retry hardening

    SooryaKiranKakkarayil
    3b835f5

  1. Day 3 complete: fix url_reputation_agent output_schema+tools bug, fix eval harness suffix bug, switch to gemini-3.1-flash-lite, all 9 EDD cases + full smoke test pass

    SooryaKiranKakkarayil
    e2329f0

  1. Day 1: scaffold, schemas, eval cases

    SooryaKiranKakkarayil
    bb8cf0a