Vision R1 training set with teacher-side visual cue text

[zaiquanyang]

thanks for your great work!

Will you consider releasing the Vision R1 training set with teacher-side visual cue text?

[tiankanghui]

Training data is publicly available at: https://zenodo.org/records/20363889

[zaiquanyang]

thanks for your feedback. I have a new issue about the geo3k_train_R.parquet and geo3k_train_S.parquet. What difference between them?

[tiankanghui]

thanks for your feedback. I have a new issue about the geo3k_train_R.parquet and geo3k_train_S.parquet. What difference between them?

Hi, R represents the answer-based privileged information, and S denotes the visual cues. Please refer to Fig. 1 in the paper for more details.

[zaiquanyang]

thanks~