Datasets:

VideoSearchR1
/

charades-stage1_data

language:
  - en
license: apache-2.0
task_categories:
  - text-retrieval
  - video-text-to-text
  - visual-question-answering
pretty_name: VideoSearch-R1 Charades-STA
configs:
  - config_name: charades_sta_stage1_cold_start
    data_files:
      - split: train
        path: Charades_STA_Stage1_ColdStart.jsonl

VideoSearch-R1 Charades-STA

This repository contains the prepared Charades-STA artifacts used by VideoSearch-R1.

Paper: VideoSearch-R1: Iterative Video Retrieval and Reasoning via Soft Query Refinement
Project Page: https://mlvlab.github.io/VideoSearch-R1/
Repository: https://github.com/mlvlab/VideoSearch-R1

Stage 1 Cold Start SFT

Charades_STA_Stage1_ColdStart.jsonl is the Stage 1 cold-start SFT dataset used to train the VideoSearch-R1 verifier/reasoner on Charades-STA. Each row keeps the ShareGPT-style messages format:

system: task instruction and output schema
user: query plus retrieved video placeholder
assistant: reasoning trace and final match decision

The original training-compatible JSON array is also retained at sft_data/train_oneturn.json. Video paths in each row refer to the prepared training video features under train/video_npy_with_meta.