Vision-and-language navigation (VLN) relies on retaining and using relevant observations. Yet growing histories introduce redundancy, and retained information may remain underused. We introduce ReInsBench, a diagnostic benchmark that examines history utilization through controlled continuation trials. Its results show...
Visual token pruning for multimodal large language models (MLLMs) appears highly effective, with existing methods retaining over 95% of unpruned performance on common VQA benchmarks using a quarter of the visual tokens or fewer. However, our extensive evaluation of ten methods across three MLLM families and five abilit...
Pei-Ran Yin, Hao Wu, Zi-Hao Yang et al.· 1 citation
Multi-view representations enable pretrained image and video vision-language models to understand 3D scenes, but their long visual token sequences make inference costly. Redundancy arises both from repeated observations of the same scene regions across views and from processing all visual tokens at every LLM layer, reg...
Yong-Xing Zheng, Hao Wu, Fang Li et al.· 2 citations
Multimodal large language models (MLLMs) incur substantial inference costs when processing long visual-textual sequences. While existing operation compression methods exploit modality-level redundancy, they largely treat computation within attention heads and shared feed-forward network (FFN) channels as unified units,...
Xu-Dong Wang, Hao Wu, Hao-Zhe Hu et al.· 1 citation
Results show that separating reusable schema encoding from selective resource access substantially reduces agentic inference costs with limited effectiveness loss.
Yichu Fang, Si-Tong Wei, Hao-Zhe Hu et al.· 3 citations
WIDE is presented, the first end-to-end differentiable token-level dynamic width pruning framework designed for both prefill and decode scenarios, and a pruning--kernel co-design framework that decomposes dynamic sparsity acceleration into mask reordering, hardware-agnostic block-level skipping, and hardware-dependent...