Research map / Vision-language and multimodal / Human and embodied modeling
Autoregressive Visual Tracking
Xing Wei, Yifan Bai, Yongchao Zheng, Dahu Shi, Yihong Gong
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 |
|---|---|---|---|
| Use autoregressive trajectory prompts instead of a per-frame model (ablation) | ARTrack · GOT-10k test | AO: 71.6% to 73.1% | Improved |
| Expand vocabulary representation range from [1×] to [2×] the search region (ablation) | ARTrack · GOT-10k test | AO: 73.1% vs. 73.5% | Improved |
| Use 800/[2 × 256] rather than 512/[2 × 256] bins per pixel (ablation) | ARTrack · GOT-10k test | AO: 71.8 -> 73.5 | Improved |
| Increase bins per pixel from 800/[2 × 256] to 1200/[2 × 256] (ablation) | ARTrack · GOT-10k test | AO: 73.5 -> 72.6 | Hurt |
| Increase bins per pixel from 800/[2 × 256] to 1600/[2 × 256] (ablation) | ARTrack · GOT-10k test | AO: 73.5 -> 71.9 | Hurt |
| Combine CE and SIoU losses rather than use either alone (ablation) | ARTrack · GOT-10k test | AO: 72.6 -> 73.5; 72.2 -> 73.5 | Improved |
| Use SIoU loss without CE loss (ablation) | ARTrack · GOT-10k test | SR0.75: 69.5 -> 67.6 | Hurt |
| Decouple and stack self- and cross-attention layers (ablation) | ARTrack · GOT-10k test | AO: 73.5 -> 73.2 | Hurt |
| Decouple and stack self- and cross-attention layers (ablation) | ARTrack · GOT-10k test | FPS: 26 -> 45 | Improved |
| Remove AR(0) pre-training (ablation) | ARTrack · LaSOT test | AUC: 70.4 -> 69.2 | Hurt |
Closest papers by meaning
- Adaptive Capacity Autoregressive Visual Tracking (CVPR 2026 · similarity 0.86)
- Autoregressive Sequential Pretraining for Visual Tracking (CVPR 2025 · similarity 0.86)
- ARTrackV2: Prompting Autoregressive Tracker Where to Look and How to Describe (CVPR 2024 · similarity 0.85)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.83)
- Autoregressive Queries for Adaptive Tracking with Spatio-Temporal Transformers (CVPR 2024 · similarity 0.83)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.82)
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.
