Papers › TACo: Token-aware Cascade Contrastive Learning for Video-Text Alignment

TACo: Token-aware Cascade Contrastive Learning for Video-Text Alignment

23 Aug 2021ICCV 2021 10arXiv:2108.09980archive 2025-07-28

Jianwei Yang, Yonatan Bisk, Jianfeng Gao

Contrastive learning has been widely used to train transformer-based vision-language models for video-text alignment and multi-modal representation learning. This paper presents a new algorithm called Token-Aware Cascade contrastive learning (TACo) that improves contrastive learning using two novel techniques. The first is the token-aware contrastive loss which is computed by taking into account the syntactic classes of words. This is motivated by the observation that for a video-text pair, the content words in the text, such as nouns and verbs, are more likely to be aligned with the visual contents in the video than the function words. Second, a cascade sampling method is applied to generate a small set of hard negative examples for efficient loss estimation for multi-modal fusion layers. To validate the effectiveness of TACo, in our experiments we finetune pretrained models for a set of downstream tasks including text-video retrieval (YouCook2, MSR-VTT and ActivityNet), video action step localization (CrossTask), video action segmentation (COIN). The results show that our models attain consistent improvements across different experimental settings over previous methods, setting new state-of-the-art on three public text-video retrieval benchmarks of YouCook2, MSR-VTT and ActivityNet.

PaperPDFConference PDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Action SegmentationContrastive LearningRepresentation LearningRetrievalTemporal Action LocalizationVideo RetrievalZero-Shot Video Retrieval

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Action Segmentation COIN TACo Frame accuracy 68.4 #6 of 9 Archive leaderboard report
Temporal Action Localization CrossTask TACo Recall 42.5 #3 of 7 Archive leaderboard report
Video Retrieval ActivityNet TACo text-to-video Median Rank 3.0 #26 of 31 Archive leaderboard report
Video Retrieval ActivityNet TACo text-to-video R@1 30.4 #26 of 31 Archive leaderboard report
Video Retrieval ActivityNet TACo text-to-video R@5 61.2 #26 of 31 Archive leaderboard report
Video Retrieval ActivityNet TACo text-to-video R@50 93.4 #26 of 31 Archive leaderboard report
Video Retrieval MSR-VTT TACo text-to-video Median Rank 5 #30 of 40 Archive leaderboard report
Video Retrieval MSR-VTT TACo text-to-video R@1 24.8 #30 of 40 Archive leaderboard report
Video Retrieval MSR-VTT TACo text-to-video R@10 64.0 #30 of 40 Archive leaderboard report
Video Retrieval MSR-VTT TACo text-to-video R@5 52.1 #30 of 40 Archive leaderboard report
Video Retrieval MSR-VTT-1kA TACo text-to-video Median Rank 4 #51 of 63 Archive leaderboard report
Video Retrieval MSR-VTT-1kA TACo text-to-video R@1 28.4 #51 of 63 Archive leaderboard report
Video Retrieval MSR-VTT-1kA TACo text-to-video R@10 71.2 #51 of 63 Archive leaderboard report
Video Retrieval MSR-VTT-1kA TACo text-to-video R@5 57.8 #51 of 63 Archive leaderboard report
Video Retrieval YouCook2 TACo text-to-video Median Rank 4 #5 of 16 Archive leaderboard report
Video Retrieval YouCook2 TACo text-to-video R@1 29.6 #5 of 16 Archive leaderboard report
Video Retrieval YouCook2 TACo text-to-video R@10 72.7 #5 of 16 Archive leaderboard report
Video Retrieval YouCook2 TACo text-to-video R@5 59.7 #5 of 16 Archive leaderboard report
Zero-Shot Video Retrieval MSR-VTT TACo text-to-video R@1 9.8 #38 of 41 Archive leaderboard report
Zero-Shot Video Retrieval MSR-VTT TACo text-to-video R@10 33.4 #38 of 41 Archive leaderboard report
Zero-Shot Video Retrieval MSR-VTT TACo text-to-video R@5 25.0 #38 of 41 Archive leaderboard report
Zero-Shot Video Retrieval YouCook2 TACo text-to-video Mean Rank 8 #5 of 9 Archive leaderboard report
Zero-Shot Video Retrieval YouCook2 TACo text-to-video R@1 19.9 #5 of 9 Archive leaderboard report
Zero-Shot Video Retrieval YouCook2 TACo text-to-video R@10 55.7 #5 of 9 Archive leaderboard report
Zero-Shot Video Retrieval YouCook2 TACo text-to-video R@5 43.2 #5 of 9 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Contrastive Learning

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections