Research map / Video / Video-language understanding
X-Pool: Cross-Modal Language-Video Attention for Text-Video Retrieval
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace text-agnostic frame pooling with text-conditioned cross-modal attention | CLIP-based text-video retrieval · MSR-VTT-9K | R@1: 44.5 -> 46.9 | Helped |
| replace text-agnostic frame pooling with text-conditioned cross-modal attention | CLIP-based text-video retrieval · MSR-VTT-7K | R@1: 42.1 -> 43.9 | Helped |
| replace text-agnostic frame pooling with text-conditioned cross-modal attention | CLIP-based text-video retrieval · MSVD | R@1: 46.2 -> 47.2 | Helped |
| replace text-agnostic frame pooling with text-conditioned cross-modal attention | CLIP-based text-video retrieval · LSMDC | R@1: 22.6 -> 25.2 | Helped |
| add text-conditioned top-k frame pooling instead of mean-pooling (ablation) | CLIP retrieval on MSR-VTT · MSR-VTT | R@1: up to 6% relative improvement | Helped |
| increase scene transitions in videos | mean-pooling retrieval · MSR-VTT augmented test set | Median Rank: 2 to 46 | Hurt |
| increase scene transitions in videos | X-Pool retrieval · MSR-VTT augmented test set | Median Rank: 2 to 9 | Hurt |
| use 6 sampled frames rather than the default 12 (ablation) | X-Pool · MSR-VTT-9K | R@1 | Hurt |
Closest papers by meaning
- DGL: Dynamic Global-Local Prompt Tuning for Text-Video Retrieval (AAAI 2024 · similarity 0.89)
- Bridging Information Asymmetry in Text-video Retrieval: A Data-centric Approach (ICLR 2025 · similarity 0.89)
- Learning Dynamic Similarity by Bidirectional Hierarchical Sliding Semantic Probe for Efficient Text Video Retrieval (AAAI 2025 · similarity 0.88)
- Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension (NeurIPS 2025 · similarity 0.88)
- Mitigating Semantic Collapse in Partially Relevant Video Retrieval (NeurIPS 2025 · similarity 0.88)
- Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning (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.