Blog
2 weeks ago
Phase 2 Calibration: Fixing Gating and Reward Scoring Together
Why per-category OOD thresholds and reward normalization were critical to Phase 2 calibration, routing stability, and reliable downstream scoring.
Source: HackerNoon →Why per-category OOD thresholds and reward normalization were critical to Phase 2 calibration, routing stability, and reliable downstream scoring.
Source: HackerNoon →