Post-hoc Reward Calibration: A Case Study on Length Bias

2025-01-22 · ICLR 2025 Poster · anchor

Findings