Research map / Video / Video-language understanding
Towards One-to-Many Temporal Grounding
Qi Xu, Tan Yue, Shihao Chen, Jiahao Meng, Anran Wang, Shunping Ji, Hao Fei, Xiangtai Li
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 SFT on OMTG dataset | Qwen3-VL-4B · OMTG Bench | EtF1: 0.21 -> 34.81 | Improved |
| add GRPO RL stage after SFT (ablation) | Qwen3-VL-4B SFT · OMTG Bench | EtF1: 34.81 -> 43.65 | Improved |
| add GRPO RL stage after SFT (ablation) | Qwen3-VL-4B SFT · OMTG Bench | C-Acc: 44.06 -> 55.63 | Improved |
| add C-Acc reward to tIoU reward (ablation) | OMTG RL · OMTG Bench | EtF1 gain over SFT: +0.74 -> +5.87 | Improved |
| add caption reward (with length penalty) (ablation) | OMTG RL · OMTG Bench | EtF1 gain over SFT: +5.87 -> +8.84 | Improved |
| add caption reward (with length penalty) (ablation) | OMTG RL · OMTG Bench | C-Acc gain over SFT: +7.50 -> +11.57 | Improved |
| add tIoU reward only (ablation) | OMTG RL · OMTG Bench | C-Acc gain over SFT: +0.31 | No clear effect |
| add data quality control (recall check and query refinement) to SFT data (ablation) | OMTG SFT · OMTG Bench | EtF1: 29.73 -> 34.81 | Improved |
| add data quality control to RL data (ablation) | OMTG RL · OMTG Bench | EtF1: 32.55 -> 43.65 | Improved |
| add RL on OMTG data only (ablation) | OMTG-4B · Charades (OOTG) | R1@0.5: 56.9 -> 58.3 | Improved |
| add RL on OMTG data only (ablation) | OMTG-4B · ActivityNet (OOTG) | R1@0.5: 50.2 -> 54.6 | Improved |
| add RL on OMTG data only (ablation) | OMTG-4B · QVHighlights (OOTG) | R1@0.5: 65.4 -> 70.4 | Improved |
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.
- Timelens-8B — TimeLens: Rethinking Video Temporal Grounding with Multimodal LLMs (CVPR 2026) · better on 1 of 1 shared result
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
- OmniVTG: A Large-Scale Dataset and Training Paradigm for Open-World Video Temporal Grounding (CVPR 2026 · similarity 0.93)
- Training-free Video Temporal Grounding using Large-scale Pre-trained Models (ECCV 2024 · similarity 0.91)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.91)
- Temporal Sentence Grounding with Relevance Feedback in Videos (NeurIPS 2024 · similarity 0.90)
- VTG-LLM: Integrating Timestamp Knowledge into Video LLMs for Enhanced Video Temporal Grounding (AAAI 2025 · similarity 0.89)
- VideoGrounding-DINO: Towards Open-Vocabulary Spatio-Temporal Video Grounding (CVPR 2024 · 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.
