Megadose AI progress, ranked and analyzed.

3DZip: Spatial-Aware Feature Diversity-Guided Token Compression for 3D Question Answering

· HF Daily Papers ·
3DZip cuts 3D VLM scene tokens to 128 while keeping 94.7% of original QA performance.

The paper targets the token overload created when 3D vision-language models project 2D features into world coordinates. Its compression pipeline uses voxelization, feature-diversity anchor selection, and spatially constrained merging to preserve geometric structure. Across three 3D question-answering benchmarks, it beats existing compression methods and reports 1.92x faster inference. Accepted to ECCV 2026. HF Daily Papers' note

score 4

Categories: Research