Research map / Video / Video-language understanding
HiVLP: Hierarchical Interactive Video-Language Pre-Training
Bin Shao, Jianzhuang Liu, Renjing Pei, Songcen Xu, Peng Dai, Juwei Lu, Weimian Li, Youliang Yan
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 multi-level visual contrastive (MVC) loss, global-to-global (ablation) | HiVLP baseline · MSR-VTT zero-shot (300k pairs) | R@10: 43.5 -> 44.2 | Improved |
| use global-to-local contrastive loss instead of global-to-global (ablation) | HiVLP · MSR-VTT zero-shot (300k pairs) | R@10: 44.2 -> 44.9 | Improved |
| use three scales of G2L contrastive loss instead of one (ablation) | HiVLP · MSR-VTT zero-shot (300k pairs) | R@10: 44.9 -> 45.3 | Improved |
| inject hierarchical features (1/32,1/16,1/8) instead of single-scale 1/8 (ablation) | HiVLP · MSR-VTT zero-shot (300k pairs) | R@1: 14.8 -> 17.4 | Improved |
| scale image-text pairs from 3M to 114M with same video-text pairs (ablation) | HiVLP · MSR-VTT zero-shot | R@1: 26.4 -> 43.5 | Improved |
| pre-train on CC3M image-text pairs instead of HD-VILA video-text pairs (ablation) | HiVLP · MSR-VTT zero-shot | R@1: 14.4 -> 17.4 | Improved |
| HiVLP (5.5M pairs) versus CLIP4Clip on fine-tuned MSR-VTT retrieval | HiVLP · MSR-VTT fine-tuned | R@1: 41.1 vs 44.5 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- SwinBERT — SwinBERT: End-to-End Transformers With Sparse Attention for Video Captioning (CVPR 2022) · better on 1 of 2 shared results
- X-Pool+ — X-Pool: Cross-Modal Language-Video Attention for Text-Video Retrieval (CVPR 2022) · 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
- Advancing High-Resolution Video-Language Representation With Large-Scale Video Transcriptions (CVPR 2022 · similarity 0.90)
- Prompting Visual-Language Models for Efficient Video Understanding (ECCV 2022 · similarity 0.89)
- Expanding Language-Image Pretrained Models for General Video Recognition (ECCV 2022 · similarity 0.89)
- Probing Visual Language Priors in VLMs (ICML 2025 · similarity 0.87)
- Enhancing Fine-Grained Vision-Language Pretraining with Negative Augmented Samples (AAAI 2025 · similarity 0.87)
- MMRL: Multi-Modal Representation Learning for Vision-Language Models (CVPR 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.
