Research map / Video / Video-language understanding
Foresee-to-Ground: From Predictive Temporal Perception to Evidence-Driven Reasoning for Video Temporal Grounding
Zelin Zheng, Xinyan Liu, Ruixin Li, Antoni B. Chan, Guorong Li, Qingming Huang, Laiyun Qing
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 conventional fine-tuning with F2G fine-tuning | Qwen3-VL-8B · Charades-STA (zero-shot) | R@0.7: 21.6 -> 25.7 | Improved |
| replace conventional fine-tuning with F2G fine-tuning | Qwen3-VL-8B · ActivityNet-Captions | mIoU: 40.8 -> 45.7 | Improved |
| replace conventional fine-tuning with F2G fine-tuning | Qwen3-VL-8B · QVHighlights (zero-shot) | HIT@1: 36.8 -> 45.6 | Improved |
| apply F2G-FT to other Video-LLM backbones | LLaVA-NeXT-7B · ActivityNet-Captions | mIoU: 12.6 -> 23.3 | Improved |
| apply F2G-FT to other Video-LLM backbones | Qwen2.5-VL-7B · Charades-STA | R@0.7: 10.3 -> 14.2 | Improved |
| add Stage-1 predictive pretraining of the temporal module (ablation) | F2G · ActivityNet-Captions | mIoU: 41.8 -> 45.7 | Improved |
| add SIGReg regularizer in Stage-1 (ablation) | F2G · ActivityNet-Captions | R@0.7: 26.8 -> 28.4 | Improved |
| add ID citation (Identify step) before timestamps (ablation) | F2G · Charades-STA (zero-shot) | R@0.7: 21.5 -> 25.7 | Improved |
| add segment visual evidence tokens Pk to evidence units (ablation) | F2G · ActivityNet-Captions | R@0.7: 23.5 -> 28.4 | Improved |
| increase Top-K from 4 to 8 (ablation) | F2G · ActivityNet-Captions | R@0.7: 25.1 -> 28.4 | Improved |
| increase SEE queries from 4 to 8 (ablation) | F2G · ActivityNet-Captions | mIoU: 45.7 -> 45.2 | No clear effect |
| fine-tune with F2G on VTG data | Qwen3-VL-8B · VideoMME (w/o subtitles) | overall accuracy: 71.4 -> 70.9 | No clear effect |
Closest papers by meaning
- TRACE: Temporal Grounding Video LLM via Causal Event Modeling (ICLR 2025 · similarity 0.89)
- VTG-LLM: Integrating Timestamp Knowledge into Video LLMs for Enhanced Video Temporal Grounding (AAAI 2025 · similarity 0.88)
- Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- Motion-Grounded Video Reasoning: Understanding and Perceiving Motion at Pixel Level (CVPR 2025 · similarity 0.88)
- Before It Fades: Reinforcing Temporal Representations at Inference Time in VideoLLMs (NeurIPS 2026 · similarity 0.88)
- Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding (NeurIPS 2025 · 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.
