Research map / Video / Video-language understanding
Training-free Video Temporal Grounding using Large-scale Pre-trained Models
Minghang Zheng, Xinhao Cai, Qingchao Chen, Yuxin Peng, Yang Liu
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 the VLM localizer with dynamic and static scoring over the naive baseline (ablation) | TFVTG · Charades-STA | R@0.5: 42.32 -> 48.01 | Improved |
| add LLM prompting alone without filtering and integration (ablation) | naive baseline · Charades-STA | R@0.7: 18.91 -> 18.56 | No clear effect |
| add LLM prompting with filtering and integration (ablation) | naive baseline · Charades-STA | mIoU: 31.61 -> 33.07 | Improved |
| add LLM prompting to VLM localizer without filtering (ablation) | TFVTG · Charades-STA | mIoU: 43.37 -> 42.76 | Hurt |
| add filtering and integration on top of LLM + VLM localizer (ablation) | TFVTG · Charades-STA | R@0.7: 21.94 -> 24.32 | Improved |
| add dynamic scoring to the naive baseline (ablation) | VLM localizer · Charades-STA | mIoU: 31.61 -> 41.68 | Improved |
| add static scoring to the naive baseline (ablation) | VLM localizer · Charades-STA | mIoU: 31.61 -> 41.81 | Improved |
| use image-text BLIP-2 instead of video-text VLMs (InterVideo, ViCLIP) (ablation) | VLM localizer · Charades-STA | mIoU: 40.72 -> 43.37 | Improved |
| use stronger LLM (GPT-4 Turbo vs GPT-3.5 Turbo) (ablation) | TFVTG · Charades-STA | R@0.7: 23.11 -> 24.32 | Improved |
| add order and relation constraints from LLM (ablation) | Filtering and integration · Charades-STA | R@0.5: 42.32 -> 44.12 | Improved |
| use training-free method on cross-dataset transfer versus supervised models | TFVTG · ActivityNet to Charades-STA | R@1 R@0.5: 21.45 -> 49.97 | Improved |
Closest papers by meaning
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.93)
- ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videos (CVPR 2025 · similarity 0.92)
- Unleashing the Potential of Multimodal LLMs for Zero-Shot Spatio-Temporal Video Grounding (NeurIPS 2025 · similarity 0.92)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.91)
- Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding (NeurIPS 2025 · similarity 0.91)
- VideoGrounding-DINO: Towards Open-Vocabulary Spatio-Temporal Video Grounding (CVPR 2024 · similarity 0.91)
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.
