search

THUDM / VisualAgentBench

Towards Large Multimodal Models as Visual Foundation Agents
Apache License 2.0
106 stars 1 forks source link

When will the training datasets be released? #3

Open zzjchen opened 2 months ago

zzjchen commented 2 months ago

Thanks for this great work! When will the datasets be released, especially the training trajectory? Could you release the datasets in very near future?

Xiao9905 commented 2 months ago

@zzjchen Hi, thanks for your interest in VAB. Our repo is still under construction, and we expect to first accomplish the quick start for evaluation. The training set is expected to be ready and released in about 1 month.

zzjchen commented 2 months ago

Thanks. I'm looking forward to it!

githubzuoyi commented 1 month ago

Thanks for this great work! Is there any plan to release the training data recently? We plan to use the VAB testing results as evaluation content.

davidkaczer commented 1 month ago

Seconding this issue, as the paper presents the curated training trajectories as one of the 3 main contributions. It would be great for the community to have these publicly available.