Research map / Vision-language and multimodal / Cross-modal learning
Fine-Grained Video-Text Retrieval With Hierarchical Graph Reasoning
Shizhe Chen, Yida Zhao, Qin Jin, Qi Wu
Check 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 |
|---|---|---|---|
| decompose matching into event, action, and entity levels and fuse their scores (ablation) | HGR · MSR-VTT | rsum: 57.6 -> 71.9 | Improved |
| add graph attention instead of averaging neighboring nodes (ablation) | HGR · MSR-VTT text-to-video | R@10: 35.6 -> 36.5 | Improved |
| add graph attention instead of averaging neighboring nodes (ablation) | HGR · MSR-VTT video-to-text | R@10: 47.1 -> 48.8 | Improved |
| add semantic-role awareness to graph reasoning (ablation) | HGR · MSR-VTT text-to-video | R@1: 9.1 -> 9.2 | Improved |
| add semantic-role awareness to graph reasoning (ablation) | HGR · MSR-VTT video-to-text | R@1: 14.2 -> 15.0 | Improved |
| add hierarchical video embeddings (ablation) | HGR · MSR-VTT | rsum: 167.9 -> 172.4 | Improved |
| use HGR instead of Dual Encoding | video-text retrieval · MSR-VTT test | rsum: 148.6 -> 172.4 | Improved |
| use HGR instead of PVSE | video-text retrieval · TGIF text-to-video | R@1: 2.3 -> 4.5 | Improved |
| use HGR instead of VSE++ | video-text retrieval · VATEX text-to-video | R@1: 33.7 -> 35.1 | Improved |
| train on MSR-VTT and evaluate HGR on an unseen dataset | HGR · YouTube2Text text-to-video | R@1: 13.8 -> 16.4 | Improved |
| use HGR rather than Dual Encoding for switching-role discrimination | fine-grained binary selection · YouTube2Text switch roles | accuracy: 71.92 -> 69.48 | Hurt |
| use HGR rather than VSE++ for incomplete-event discrimination | fine-grained binary selection · YouTube2Text incomplete events | accuracy: 78.79 -> 82.04 | Improved |
Closest papers by meaning
- Learning Fine-Grained Representations through Textual Token Disentanglement in Composed Video Retrieval (ICLR 2025 · similarity 0.90)
- Learning Dynamic Similarity by Bidirectional Hierarchical Sliding Semantic Probe for Efficient Text Video Retrieval (AAAI 2025 · similarity 0.88)
- Stitching Segments and Sentences towards Generalization in Video-Text Pre-training (AAAI 2024 · similarity 0.88)
- Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding (NeurIPS 2025 · similarity 0.88)
- Mitigating Semantic Collapse in Partially Relevant Video Retrieval (NeurIPS 2025 · similarity 0.87)
- Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning (AAAI 2024 · 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.
