Research map / Video / Video-language understanding
DART: Difficulty-Adaptive Routing for Zero-Shot Video Temporal Grounding
zhengbo zhang, Mark H. Huang, Zhigang Tu, Ming-Hsuan Yang
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 query-conditioned DPP keyframe selection (vs uniform sampling) (ablation) | DART · Charades-STA (IID) | mIoU: 44.70 -> 48.93 | Improved |
| add adaptive keyframe budget (vs fixed K=8) (ablation) | DART · Charades-STA (IID) | mIoU: 47.26 -> 48.93 | Improved |
| add Temporal Markup Prompting (vs generic chain-of-thought) (ablation) | DART · Charades-STA (IID) | mIoU: 46.13 -> 48.93 | Improved |
| add Slow path reasoning (vs Fast only) (ablation) | DART · Charades-STA (IID) | mIoU: 44.34 -> 47.53 | Improved |
| add spectral-entropy routing (vs Slow only) (ablation) | DART · Charades-STA (IID) | mIoU: 47.53 -> 48.93 | Improved |
| add spectral-entropy routing (vs Fast only) (ablation) | DART · Charades-STA (IID) | mIoU: 44.34 -> 48.93 | Improved |
| use DPP instead of Top-k + NMS keyframe selection (ablation) | DART · Charades-STA (IID) | mIoU: 47.38 -> 48.93 | Improved |
| use DPP instead of K-medoids (ablation) | DART · Charades-STA (IID) | mIoU: 45.73 -> 48.93 | Improved |
| use DPP instead of uniform + re-rank (ablation) | DART · Charades-STA (IID) | mIoU: 46.52 -> 48.93 | Improved |
| add adaptive routing to Slow-only inference (ablation) | DART · Charades-STA inference | time per query: 5.1s -> 3.9s | Improved |
| add routing: Fast-path for simple queries (ablation) | DART · Charades-STA (IID) | mIoU: 47.53 -> 48.93 | Improved |
| DART vs best prior zero-shot (TAG) in IID | DART · ActivityNet Captions (IID) | mIoU: 36.55 -> 39.89 | Improved |
Closest papers by meaning
- Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal Grounding (ECCV 2026 · similarity 0.91)
- SE-DETR: Explicit Semantic Exploration for Generalizability and Distinguishability in Video Temporal Grounding (ECCV 2026 · similarity 0.91)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.90)
- Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding (NeurIPS 2025 · similarity 0.90)
- Unleashing the Potential of Multimodal LLMs for Zero-Shot Spatio-Temporal Video Grounding (NeurIPS 2025 · similarity 0.89)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.89)
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.
