Research map / Video / Video-language understanding
TACo: Token-Aware Cascade Contrastive Learning for Video-Text Alignment
Jianwei Yang, Yonatan Bisk, Jianfeng Gao
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 token-level loss L2 (noun+verb) on top of L1+L3 with cascade (ablation) | TACo · YouCook2 R@1 | R@1: 15.0 -> 15.8 | Improved |
| add token-level loss L2 (noun+verb) on top of L1+L3 with cascade (ablation) | TACo · MSR-VTT split1 R@1 | R@1: 23.7 -> 24.5 | Improved |
| add cascade hard negative sampling to L1+L3 (ablation) | TACo · YouCook2 R@1 | R@1: 13.9 -> 15.0 | Improved |
| add cascade hard negative sampling to L1+L3 (ablation) | TACo · MSR-VTT split1 R@1 | R@1: 22.5 -> 23.7 | Improved |
| combine early-stage L1 with later-stage L3 instead of L3 only (ablation) | TACo · YouCook2 R@1 | R@1: 13.3 -> 13.9 | Improved |
| token loss on determiners and adpositions instead of none (ablation) | TACo · YouCook2 R@1 | R@1: 15.0 -> 14.7 | Hurt |
| token loss on nouns only instead of none (ablation) | TACo · YouCook2 R@1 | R@1: 15.0 -> 15.4 | Improved |
| token loss on verbs only instead of none (ablation) | TACo · YouCook2 R@1 | R@1: 15.0 -> 15.3 | Improved |
| use nouns+verbs instead of nouns only as tokens of interest (ablation) | TACo · YouCook2 R@1 | R@1: 15.4 -> 15.8 | Improved |
| add video encoder layers (0 to 4) (ablation) | TACo · YouCook2 R@1 (R-152+S3D-HM) | R@1: 14.0 -> 16.4 | Improved |
| use cascade sampling on later-stage alignment instead of early+later stage losses only (ablation) | TACo · YouCook2 later-stage R@1 | R@1: 13.6 -> 14.5 | Improved |
| use TACo losses instead of baseline loss | TACo · YouCook2 R@1 (R-152+S3D-HM) | R@1: 13.3 -> 15.8 | Improved |
Later papers that compared with this method
7 later accepted papers report TACO as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.
- Image Guides Images: Consistent Video Amodal Completion with Rectified In-Context Exemplar Guidance (CVPR 2026 · better on 9 of 10 shared results)
- Learning Semantic Alignment with Global Modality Reconstruction for Video-Language Pre-training towards Retrieval (AAAI 2023 · better on 3 of 3 shared results)
- Temporal Alignment Networks for Long-Term Video (CVPR 2022 · better on 1 of 1 shared result)
- Long-Form Video-Language Pre-Training with Multimodal Temporal Contrastive Learning (NeurIPS 2022 · better on 1 of 1 shared result)
- LGDN: Language-Guided Denoising Network for Video-Language Modeling (NeurIPS 2022 · better on 1 of 1 shared result)
- Composable Augmentation Encoding for Video Representation Learning (ICCV 2021 · better on 2 of 2 shared results)
- Enhancing Self-Supervised Video Representation Learning via Multi-Level Feature Optimization (ICCV 2021 · better on 2 of 2 shared results)
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.
- ActBert — ActBERT: Learning Global-Local Video-Text Representations (CVPR 2020) · 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
- VideoCon: Robust Video-Language Alignment via Contrast Captions (CVPR 2024 · similarity 0.89)
- Rebalancing Contrastive Alignment with Bottlenecked Semantic Increments in Text-Video Retrieval (NeurIPS 2025 · similarity 0.88)
- VideoComp: Advancing Fine-Grained Compositional and Temporal Alignment in Video-Text Models (CVPR 2025 · similarity 0.87)
- Stitching Segments and Sentences towards Generalization in Video-Text Pre-training (AAAI 2024 · similarity 0.87)
- DiscoVLA: Discrepancy Reduction in Vision, Language, and Alignment for Parameter-Efficient Video-Text Retrieval (CVPR 2025 · similarity 0.86)
- Improving fine-grained understanding in image-text pre-training (ICML 2024 · 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.
