Research map / Video / Video-language understanding
Long-Form Video-Language Pre-Training with Multimodal Temporal Contrastive Learning
Yuchong Sun, Hongwei Xue, Ruihua Song, Bei Liu, Huan Yang, Jianlong Fu
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 MTC loss to global contrastive loss (ablation) | LF-VILA · ActivityNet Captions | R@1: 26.1 -> 27.8 | Improved |
| Add MTC loss to global contrastive loss (ablation) | LF-VILA · QuerYD | R@1: 55.4 -> 57.3 | Improved |
| Increase frames from 8 to 32 with one clip (ablation) | LF-VILA with global loss · ActivityNet Captions | R@1: 19.3 -> 24.3 | Improved |
| Increase pre-training clips from 1 to 4, using 8 frames (ablation) | LF-VILA with global loss · QuerYD | R@1: 38.4 -> 41.6 | Improved |
| Replace no pre-training with global-contrastive pre-training using 1 clip and 8 frames (ablation) | LF-VILA · ActivityNet Captions | R@1: 15.0 -> 19.3 | Improved |
| Replace HTWA with fixed temporal window 4 (ablation) | LF-VILA · ActivityNet Captions | R@1: 26.1 -> 25.1 | Hurt |
| Replace HTWA with fixed temporal window 32 (ablation) | LF-VILA · ActivityNet Captions | R@1: 26.1 -> 26.2 | No clear effect |
| Replace HTWA with fixed temporal window 32 (ablation) | LF-VILA · ActivityNet Captions | training time: 1.00× -> 1.49× | Hurt |
| Use LF-VILA rather than TACo | paragraph-to-video retrieval · ActivityNet Captions | R@1: 30.4 -> 35.3; 16.1% | Improved |
| Use LF-VILA rather than Frozen | paragraph-to-video retrieval · DiDeMo | R@1: 31.0 -> 35.0; 12.9% | Improved |
| Use LF-VILA rather than Frozen* | paragraph-to-video retrieval · QuerYD | R@1: 53.8 -> 69.7; 29.6% | Improved |
| Use LF-VILA rather than TeachText | paragraph-to-video retrieval · Condensed Movie | R@1: 12.08 -> 13.56; 12.3% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TACo — TACo: Token-Aware Cascade Contrastive Learning for Video-Text Alignment (ICCV 2021) · better on 1 of 1 shared result
- TeachText — TeachText: CrossModal Generalized Distillation for Text-Video Retrieval (ICCV 2021) · 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
- Unleashing Hour-Scale Video Training for Long Video-Language Understanding (NeurIPS 2025 · similarity 0.91)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.90)
- Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization (ICML 2024 · similarity 0.89)
- LanguageBind: Extending Video-Language Pretraining to N-modality by Language-based Semantic Alignment (ICLR 2024 · similarity 0.88)
- TOPA: Extending Large Language Models for Video Understanding via Text-Only Pre-Alignment (NeurIPS 2024 · similarity 0.88)
- Stitching Segments and Sentences towards Generalization in Video-Text Pre-training (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.
