Research map / Video / Video-language understanding
Variational Cross-Graph Reasoning and Adaptive Structured Semantics Learning for Compositional Temporal Grounding
Juncheng Li, Siliang Tang, Linchao Zhu, Wenqiao Zhang, Yi Yang, Tat‐Seng Chua, Fei Qing Wu, Yueting Zhuang
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 Semantic-Contextualized Learning (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 43.75 -> 45.41 | Improved |
| add Visual-Contextualized Learning (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 42.26 -> 45.41 | Improved |
| add Hierarchical Semantics Aggregation (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 44.22 -> 45.41 | Improved |
| use variational cross-graph correspondence instead of cross-modal self-attention fusion (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 41.08 -> 45.41 | Improved |
| use semantic graphs instead of raw detection results and SRL labels as features (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 12.97 -> 45.41 | Improved |
| use semantic graphs instead of raw detection results as features (ablation) | VISA · ActivityNet-CG novel-composition | R@1 IoU=0.5: 10.92 -> 31.51 | Improved |
| add VCC (variational correspondence) instead of self-attention (ablation) | VISA · ActivityNet-CG novel-composition | R@1 IoU=0.5: 27.32 -> 31.51 | Improved |
| add masked semantics modeling loss (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 45.40 -> 46.37 | Improved |
| add masked semantics modeling loss (ablation) | VISA · ActivityNet-CG novel-composition | R@1 IoU=0.5: 31.51 -> 32.09 | Improved |
| add structure-consistency loss (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 45.40 -> 46.41 | Improved |
| add semantic-adaptive margin loss (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 45.40 -> 46.70 | Improved |
| add all three ASSL losses (ablation) | VISA · Charades-CG novel-composition | R@1 IoU=0.5: 45.40 -> 47.76 | Improved |
Closest papers by meaning
- Compositional Temporal Grounding With Structured Variational Cross-Graph Correspondence Learning (CVPR 2022 · similarity 0.98)
- Investigating Compositional Challenges in Vision-Language Models for Visual Grounding (CVPR 2024 · similarity 0.88)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.87)
- VELOCITI: Benchmarking Video-Language Compositional Reasoning with Strict Entailment (CVPR 2025 · similarity 0.87)
- Multi-Pair Temporal Sentence Grounding via Multi-Thread Knowledge Transfer Network (AAAI 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.
