Stage 2 starts from the ActivityNet Stage 1 checkpoint and optimizes iterative retrieval and temporal grounding behavior with the VideoSearch-R1 training pipeline.

Usage

Project Page: <a href="https://mlvlab.github.io/VideoSearch-R1/" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">https://mlvlab.github.io/VideoSearch-R1/
Repository: <a href="https://github.com/mlvlab/VideoSearch-R1" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">https://github.com/mlvlab/VideoSearch-R1

Use with the VideoSearch-R1 codebase:

bash

bash scripts/data_construct/download_preextracted_data.bash activitynet
EVAL_GPUS=0 bash scripts/inference/inference.bash activitynet --checkpoint VideoSearchR1/activitynet-stage2

Citation

bibtex

@inproceedings{lee2026videosearchr1,
  title     = {VideoSearch-R1: Iterative Video Retrieval and Reasoning via Soft Query Refinement},
  author    = {Lee, Seohyun and Choi, Seoung and Ko, Dohwan and Kim, Jongha and Kim, Hyunwoo J.},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year      = {2026}
}