4 papers
From Token Importance to Conditional Removability: Rethinking Visual Token Pruning in Multimodal Large Language Models
Shengli He, Yongchao Liang, Roumeng He +5
Training-free visual-token pruning often uses token importance, redundancy, or related selection criteria as proxies for safe removal. We show that these signals alone do not fully…
Token Utility Is Selection-Conditioned: Coupled Selection of Prompt Context and Response Supervision for Efficient Instruction Tuning
Can Wu, Xinrui Chen, Ou Wu +1
Efficient large language model (LLM) instruction tuning requires selecting response supervision with supporting prompt context. Existing methods typically value both sides separate…
QCPruner: Query-Conditioned Population Coverage for Visual Token Pruning
Shengli He, Yongchao Liang, Roumeng He +4
The high visual-token load in multimodal large language models (MLLMs) motivates training-free pruning to reduce later-layer computation, but under a fixed budget, pruning must pre…
Spatiotemporal Gaussian representation-based dynamic reconstruction and motion estimation framework for time-resolved volumetric MR imaging (DREME-GSMR)
Jiacheng Xie, Hua-Chieh Shao, Can Wu +7
Time-resolved volumetric MR imaging that reconstructs a 3D MRI within sub-seconds to resolve deformable motion is essential for motion-adaptive radiotherapy. Representing patient a…