Our paper “Investigating Structural Pruning and Recovery Techniques for Compressing MLLMs” was accepted to the German Conference on Pattern Recognition (GCPR) 2025 as an Oral presentation.
This is joint work with Lukas Thede, Massimiliano Mancini, Wei Xu, and Zeynep Akata.
We develop pruning and recovery pipelines for LLaVA, Bunny, and InternVL, and study how much recovery data is actually needed to restore performance after pruning. The journal version was later accepted to IJCV 2026.