Research map / Label-efficient and robust learning / Test-time adaptation
Robust Test-time Video-Text Retrieval: Benchmarking and Adapting for Query Shifts
Bingqing Zhang, Zhuo Cao, Heming Du, Yang Li, Xue Li, Jiajun Liu, Sen Wang
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 HAT-VTR (hubness suppression + multi-granular TCR) to CLIP4Clip | CLIP4Clip · MSRVTT-1kA v2t, severity 5 | R@1 avg: 17.0 -> 26.2 | Improved |
| add HAT-VTR to Xpool | Xpool · MSRVTT-1kA v2t, severity 5 | R@1 avg: 20.0 -> 30.3 | Improved |
| add HAT-VTR to CLIP4Clip | CLIP4Clip · ActivityNet v2t, highest severity | R@1 avg: 12.41 -> 22.28 | Improved |
| apply existing TTA methods (Tent, READ, SAR, EATA) to CLIP4Clip | CLIP4Clip · ActivityNet v2t, highest severity | R@1 avg: 12.41 -> 10.23 (Tent), 11.07 (EATA) | Hurt |
| apply existing TTA methods (Tent, READ, SAR) to CLIP4Clip | CLIP4Clip · MSRVTT-1kA v2t, severity 5 | R@1 avg: 17.0 -> 16.8 (Tent), 17.0 (READ), 17.2 (SAR) | No clear effect |
| add TCR to CLIP4Clip | CLIP4Clip · MSRVTT->ActivityNet cross-dataset v2t | R@1: 32.64 -> 19.28 | Hurt |
| add HAT-VTR to CLIP | CLIP · zero-shot adaptation, MSRVTT v2t | R@1: 26.50 -> 35.40 | Improved |
| add HSM at posterior reranking (ablation) | HAT-VTR · HSM ablation, v2t | R@1: 23.3 -> 25.5 | Improved |
| add HSM at target selection (ablation) | HAT-VTR · HSM ablation, v2t | R@1: 23.3 -> 23.6 | Improved |
| add HSM at both target selection and reranking (ablation) | HAT-VTR · HSM ablation | avg R@1: 28.0 -> 30.1 | Improved |
| add noise-robust entropy term L_NA (ablation) | HAT-VTR adaptation loss · adaptation-loss ablation, v2t | R@1: 22.6 -> 24.1 | Improved |
| add multi-granular uniformity losses L_inter and L_intra (ablation) | HAT-VTR adaptation loss · adaptation-loss ablation, v2t | R@1: 22.6 -> 24.3 | Improved |
Closest papers by meaning
- TempMe: Video Temporal Token Merging for Efficient Text-Video Retrieval (ICLR 2025 · similarity 0.88)
- Mitigating Semantic Collapse in Partially Relevant Video Retrieval (NeurIPS 2025 · similarity 0.87)
- Text Proxy: Decomposing Retrieval from a 1-to-N Relationship into N 1-to-1 Relationships for Text-Video Retrieval (AAAI 2025 · similarity 0.87)
- DiscoVLA: Discrepancy Reduction in Vision, Language, and Alignment for Parameter-Efficient Video-Text Retrieval (CVPR 2025 · similarity 0.87)
- MV-Adapter: Multimodal Video Transfer Learning for Video Text Retrieval (CVPR 2024 · similarity 0.87)
- Bridging Information Asymmetry in Text-video Retrieval: A Data-centric Approach (ICLR 2025 · similarity 0.86)
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.
