Research map / Architectures and efficiency / Attention and transformers
HTTM: Head-wise Temporal Token Merging for Faster VGGT
Weitian Wang, Lukas Meiner, Rai Shubham, Cecilia De La Parra, Akash Kumar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adaptive outlier filtering (10%) to block-wise merging (ablation) | HTTM · NRGBD | accuracy (Chamfer): 0.240 -> 0.012 | Improved |
| 5% vs 10% outlier filtering (ablation) | HTTM · NRGBD | accuracy (Chamfer): 0.013 vs 0.012 | No clear effect |
| apply HTTM to VGGT (Q 0.20, K/V 0.30) | VGGT · 7 Scenes (stride 10) | accuracy (Chamfer): 0.019 -> 0.020 | Hurt |
| apply HTTM to VGGT | VGGT · 7 Scenes (stride 10) | time: 9.1s -> 4.3s | Improved |
| apply HTTM to VGGT | VGGT · NRGBD (stride 10) | accuracy (Chamfer): 0.010 -> 0.012 | Hurt |
| HTTM vs FastVGGT | VGGT · NRGBD (stride 10) | accuracy (Chamfer): 0.016 -> 0.012 | Improved |
| HTTM vs FastVGGT | VGGT · ETH3D | accuracy (Chamfer): 0.120 -> 0.091 | Improved |
| HTTM vs FastVGGT | VGGT · VGGT latency, 1000 frames | latency: 175.2s -> 102.8s | Improved |
| block-wise merging after temporal reordering vs global matching | HTTM vs FastVGGT · Global attention latency, 1000 frames | matching time: 2.31s -> 0.12s | Improved |
| adaptive outlier filtering increases aggregation overhead | HTTM vs FastVGGT · Global attention latency, 1000 frames | merged token aggregation time: 0.11s -> 0.41s | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FastVGGT — FastVGGT: Fast Visual Geometry Transformer (ICLR 2026) · better on 8 of 10 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token Merging (CVPR 2026 · similarity 0.90)
- VGGT: Visual Geometry Grounded Transformer (CVPR 2025 · similarity 0.89)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.89)
- RegimeVGGT: Layer-Wise Spatially Preserving Redundancy Removal for Visual Geometry Grounded Transformer (NeurIPS 2026 · similarity 0.88)
- Merge3D: Efficient 3D Multimodal LLMs via Joint 2D-3D Token Merging (CVPR 2026 · similarity 0.86)
- Good Token Hunting: A Hitchhiker's Guide to Token Selection for Visual Geometry Transformers (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
