Preprint
Sep 2026
CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction
Evaluated in a multi-teacher setting for GUI grounding and optical character recognition, CA-OPD substantially improves the Qwen3.5-0.8B baseline across all six target benchmarks, including gains of $9.50$ points on ScreenSpot-Pro and OCRBench-v2 English.
Meng-Hao Li, Lin-Jie Mu, Yin Wang et al.
· 2 citations