Research map / Video / Video-language understanding
Rethinking Weakly-supervised Video Temporal Grounding From a Game Perspective
Xiang Fang, Zeyu Xiong, Wanlong Fang, Xiaoye Qu, Chen Chen, Jianfeng Dong, Keke Tang, Pan Zhou, Yu Cheng, Daizong 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 video self-modal game (ablation) | word-level cross-modal game baseline · Charades-STA | R@1, IoU=0.7: 19.34 -> 21.25 | Improved |
| add query self-modal game (ablation) | video game + word-level · Charades-STA | R@1, IoU=0.7: 21.25 -> 22.97 | Improved |
| add phrase-level cross-modal game (ablation) | BA-Game with self games · Charades-STA | R@1, IoU=0.7: 22.97 -> 24.81 | Improved |
| add sentence-level cross-modal game (ablation) | full model · Charades-STA | R@1, IoU=0.7: 24.81 -> 26.46 | Improved |
| add KLD soft supervision to the game header (ablation) | game headers · Charades-STA | R@1, IoU=0.7: 23.21 -> 26.46 | Improved |
| use KLD instead of MSE or CE for soft supervision (ablation) | game headers · Charades-STA | R@1, IoU=0.7: 24.73 (CE) -> 26.46 | Improved |
| use convolution layer plus self-attention header (ablation) | game headers · Charades-STA | R@1, IoU=0.7: 25.28 (LL) -> 26.46 | Improved |
| add reconstruction loss (ablation) | full model · Charades-STA | R@1, IoU=0.7: 24.35 -> 26.46 | Improved |
| replace Banzhaf index with Shapley index (ablation) | game framework · Charades-STA | R@1, IoU=0.7: 26.46 -> 26.28 | No clear effect |
| replace proposal-based matching with frame-wise game scores | WS-VTG · Charades-STA | R@5, IoU=0.5: 84.71 -> 88.93 vs CPL | Improved |
| replace proposal-based matching with game-based frame scoring in cross-dataset transfer | WS-VTG · Cross-dataset C->A | R@1, IoU=0.5: 11.59 -> 20.37 vs CPL | Improved |
| game-based training with coalition sampling versus CPL | training cost · training speed per sample | seconds: 0.27s -> 0.94s | Hurt |
Closest papers by meaning
- G2L: Semantically Aligned and Uniform Video Grounding via Geodesic and Game Theory (ICCV 2023 · similarity 0.91)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- Contextual Self-paced Learning for Weakly Supervised Spatio-Temporal Video Grounding (ICLR 2025 · similarity 0.88)
- Local-Global Multi-Modal Distillation for Weakly-Supervised Temporal Video Grounding (AAAI 2024 · similarity 0.88)
- Learning Multi-Scale Video-Text Correspondence for Weakly Supervised Temporal Article Gronding (AAAI 2024 · similarity 0.88)
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.
