Ming-Jun Liu

Selective Transfer of RL Updates for Visual Reasoning

This work isolates the RL-stage update, retains its dominant matrix-wise directions with magnitude preservation, and transfers them to the language modules of a VLM, which improves full-update interpolation in 12 of 15 comparisons, including an 8.55 percentage-point MathVision gain on the Qwen recipient.

MMNet: multi-scale VSSM and masked image reconstruction network for robust crowd counting