---
title: "Discussion on \"Why Temporal Alignment Matters in Video-Language Datasets\" | Hashnode"
description: "Discussion on \"Why Temporal Alignment Matters in Video-Language Datasets\". A video and a caption are not automatically a useful training pair.\nFor many multimodal tasks, the model needs to understand not only what appears in a video, but when a specific event, action, or sce"
url: "https://hashnode.com/posts/why-temporal-alignment-matters-in-video-language-datasets/6ae2e28d865bda097de6442c"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Why Temporal Alignment Matters in Video-Language Datasets](https://thordata-notes.hashnode.dev/why-temporal-alignment-matters-in-video-language-datasets?utm_source=hashnode&utm_medium=feed)

A video and a caption are not automatically a useful training pair. For many multimodal tasks, the model needs to understand not only what appears in a video, but when a specific event, action, or sce

Oct 29 · 3 min read · thordata-notes.hashnode.dev  
Read the article

## Written by

[thordata](https://hashnode.com/@thordata-builds)

@thordata-builds

  
[View profile](https://hashnode.com/@thordata-builds)

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
