Video-Text Retrieval Performance After Removing Crossmodal Contrastive Losses in Large-Scale Video Encoders

This report synthesises findings from 15 peer-reviewed papers addressing the following research question: What is the effect of removing crossmodal contrastive learning auxiliary losses on the video-text retrieval performance of large-scale video encoders on the MSR-VTT benchmark. 10 claims were extracted from source literature; 10 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 9.2/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: What is the effect of removing crossmodal contrastive learning auxiliary losses on the video-text retrieval performance of large-scale video encoders on the MSR-VTT benchmark?

Autonomous literature synthesis. Automated review score: 9.2/10. Full text and citation available at Assignee Research.