Commit History

  1. Sanitize empty finding descriptions

    e3e5d23
  2. Harden reward parsing for truncated JSON

    fd84f8b

  1. Round2 hardening, campaign fixes, training quickstart, and CI gate

    92ead25
  2. Round 2 Implementation: Multi-period campaign, regulatory shocks, adversarial grading, and GRPO training infrastructure

    52f5c27

  1. Emergency fix: drop scoring limit to 2 decimal points and 0.01 / 0.99 to pass deep verification

    57d984d

  1. Fix phase 2 deep validation grader scoring out-of-bounds error

    140ce9d
  2. Refactor graders.py: add explicit partial credit comments, strict duplicate handling, and weighted false negative tracking

    e43c1bc
  3. Fix inference stdout formatting and update graders tests for strictly clamped reward constraint bounds

    21fc032
  4. fix: increase score epsilon to 0.01 for validator safety

    b4d5e6a

  1. feat: v2.0 — fraud detection task, severity grading, investigation mode, security hardening, 78 tests⁠

    126bdbd