Research map / Video / Video-language understanding
Gravitation-Driven Semantic Alignment for Text Video Retrieval
Yi Yang, Zheng Wang, Xing Xu, Jingkuan Song, Heng Tao Shen
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 semantic attraction term A to the alignment score (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | R@1: 50.5 -> 52.4 | Improved |
| add alignment penalty term B (Mahalanobis) to the alignment score (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | R@1: 51.9 -> 52.4 | Improved |
| add uncertainty regularizer term C to the alignment score (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | R@1: 52.1 -> 52.4 | Improved |
| add semantic attraction term A (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | MnR: 15.4 -> 12.9 | Improved |
| add GraviAlign loss to CLIP-ViP | CLIP-ViP ViT-B/32 · MSR-VTT T2V | R@1: 50.1 -> 52.4 | Improved |
| add GraviAlign loss to CLIP-ViP | CLIP-ViP ViT-B/16 · MSR-VTT T2V | R@1: 54.2 -> 55.8 | Improved |
| add GraviAlign loss to CLIP-ViP | CLIP-ViP ViT-B/16 · ActivityNet Captions val1 | R@1: 53.4 -> 56.0 | Improved |
| add GraviAlign loss to CLIP-ViP | CLIP-ViP ViT-B/32 · DiDeMo | R@1 / R@5: R@1 48.6 -> 50.7; R@5 77.1 -> 76.9 | Mixed |
| vary temperature T from 1.0 to 0.5 or 2.0 (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | R@1: 52.4 -> 52.3 / 52.1 | No clear effect |
| increase mass sensitivity lambda from 0.01 to 0.05 or 0.1 (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | R@1: 52.4 -> 51.9 / 52.0 | Hurt |
| change gravitational constant G from 0.1 to 0.05 or 1.0 (ablation) | GraviAlign · MSR-VTT T2V (ViT-B/32) | R@1: 52.4 -> 51.9 / 52.3 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UATVR — UATVR: Uncertainty-Adaptive Text-Video Retrieval (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Rebalancing Contrastive Alignment with Bottlenecked Semantic Increments in Text-Video Retrieval (NeurIPS 2025 · similarity 0.86)
- Mitigating Semantic Collapse in Partially Relevant Video Retrieval (NeurIPS 2025 · similarity 0.85)
- Learning Dynamic Similarity by Bidirectional Hierarchical Sliding Semantic Probe for Efficient Text Video Retrieval (AAAI 2025 · similarity 0.84)
- Rethinking Noisy Video-Text Retrieval via Relation-aware Alignment (CVPR 2025 · similarity 0.84)
- Text Is MASS: Modeling as Stochastic Embedding for Text-Video Retrieval (CVPR 2024 · similarity 0.83)
- Towards Balanced Alignment: Modal-Enhanced Semantic Modeling for Video Moment Retrieval (AAAI 2024 · similarity 0.83)
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.
