Research map / Architectures and efficiency / Attention and transformers
LongStream: Long-Sequence Streaming Autoregressive Visual Geometry
Chong Cheng, Xianda Chen, Tao Xie, Wei Yin, Weiqiang Ren, Qian Zhang, Xiaoyang Guo, Hao Wang
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 |
|---|---|---|---|
| replace absolute pose with keyframe-relative pose (ablation) | LongStream · vKITTI single sequence (ablation) | ATE: 8.043 -> 2.819 | Improved |
| add scale head (ablation) | LongStream · vKITTI single sequence (ablation) | ATE: 2.819 -> 2.645 | Improved |
| add scale head (ablation) | LongStream · vKITTI single sequence (ablation) | RPE: 0.750 -> 0.484 | Improved |
| add cache-consistent training (ablation) | LongStream · vKITTI single sequence (ablation) | ATE: 2.645 -> 0.984 | Improved |
| add periodic cache refresh (ablation) | LongStream · vKITTI single sequence (ablation) | ATE: 0.984 -> 0.115 | Improved |
| add CCT and cache refresh (ablation) | LongStream · vKITTI single sequence (ablation) | Scale error: 0.010 -> 0.035 | Hurt |
| use LongStream vs best streaming baseline | LongStream · KITTI avg | ATE: 185.89 -> 92.55 (CUT3R) | Improved |
| use LongStream vs MASt3R-SLAM on TUM reconstruction | LongStream · TUM | F1@0.25: 0.954 -> 0.673 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- FastVGGT — FastVGGT: Fast Visual Geometry Transformer (ICLR 2026) · better on 1 of 3 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
- CLoSeR: Closing the Loop for Long-Context Streaming Reconstruction (NeurIPS 2026 · similarity 0.90)
- FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry (NeurIPS 2026 · similarity 0.89)
- IGGT4D: Streaming 4D Instance-Grounded Geometry Transformer (NeurIPS 2026 · similarity 0.88)
- GHOST: Geometry-Hierarchical Online Streaming Token Eviction for Efficient 3D Reconstruction (NeurIPS 2026 · similarity 0.88)
- Rethinking the State Update Gate for Long-Sequence Recurrent 3D Reconstruction (NeurIPS 2026 · similarity 0.87)
- 3DGStream: On-the-Fly Training of 3D Gaussians for Efficient Streaming of Photo-Realistic Free-Viewpoint Videos (CVPR 2024 · similarity 0.87)
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.
