Research map / Video / Video-language understanding
Text-Visual Prompting for Efficient 2D Temporal Video Grounding
Yimeng Zhang, Xin Chen, Jinghan Jia, Sijia Liu, Ke Ding
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 both visual and text prompts (ablation) | 2D TVG base model · Charades-STA | R@1, IoU=0.5: 40.43 -> 44.39 | Improved |
| Add visual prompts alone (ablation) | 2D TVG base model · ActivityNet Captions | R@1, IoU=0.7: 19.14 -> 23.71 | Improved |
| Add text prompts alone (ablation) | 2D TVG base model · ActivityNet Captions | R@1, IoU=0.7: 19.14 -> 24.39 | Improved |
| Combine text and visual prompts rather than use either alone (ablation) | 2D TVG base model · ActivityNet Captions | R@1, IoU=0.7: 30.77% | Improved |
| Replace padding-region pixels with visual prompts rather than add them (ablation) | 2D TVG base model · Charades-STA | R@1, IoU=0.3: 61.29 -> 65.38 | Improved |
| Add temporal-distance loss to tIoU loss (ablation) | TVP · Charades-STA | R@1, IoU=0.7: 11.82 -> 16.77 | Improved |
| Add duration loss to tIoU loss (ablation) | TVP · Charades-STA | R@1, IoU=0.7: 11.82 -> 16.34 | Improved |
| Add both distance and duration losses to tIoU loss (ablation) | TVP · Charades-STA | R@1, IoU=0.7: 68.27% | Improved |
| Use a visual prompt size of 96 (ablation) | TVP visual prompts · Charades-STA | R@1, IoU=0.3: 61.29 -> 65.38 | Improved |
| Increase visual prompt size beyond 96 (ablation) | TVP visual prompts · Charades-STA | R@1, IoU=0.3: 65.38 -> 64.73 | Hurt |
| Use text prompts rather than no text prompts (ablation) | TVP text prompts · Charades-STA | R@1, IoU=0.3: 57.20 -> 65.81 | Improved |
| Increase sampled frames beyond 48 (ablation) | Base model · Charades-STA | R@1 | Hurt |
Closest papers by meaning
- Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding (NeurIPS 2025 · similarity 0.90)
- SnAG: Scalable and Accurate Video Grounding (CVPR 2024 · similarity 0.88)
- Unleashing the Potential of Multimodal LLMs for Zero-Shot Spatio-Temporal Video Grounding (NeurIPS 2025 · similarity 0.88)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.87)
- SimVG: A Simple Framework for Visual Grounding with Decoupled Multi-modal Fusion (NeurIPS 2024 · similarity 0.87)
- LaViP: Language-Grounded Visual Prompting (AAAI 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.
