Being confused by the datastructure

[YuanLi-X]

Hi! This benchmark is very meaningful! However, I am a bit confused by the secondary structure datasets. The datasets provided here are different from that on HuggingFace. Further, the number of train/test samples (the train/test files of '.csv' format) are different from the description in the paper (Table A10). Could you please give me some detailed clarification about the SSP datasets. It will be better if you can give some guidance on how to reproduce the results presented in the paper, because I do not know which file should be used when reproduced the prediction score of RNA-FM. Thanks so much!!

[murfalo]

Hello! Thank you for your feedback. I believe this may be a duplicate of another issue. In short, the difference between HuggingFace and our self-hosted files is indeed a bug, the current documentation is confusing, and we are actively working to address this. We plan to address this with clarified steps for reproduction in the next edition of RNAGym which will be released in the coming months. Thank you for your patience.

[YuanLi-X]

Thanks so much for your reply. However, I still have another question. I noticed the description: '`RMDB_dataset_<integer>.parquet` contains the data from only the ribonanza train and test set.' in the README.md file. However, the test file provided by the ribonanza did not give details about the dataset name. So how did you construct the `RMDB_dataset_<integer>.parquet files using the test set? Thanks so much again and looking forward to your reply.

…

________________________________ From: murfalo ***@***.***> Sent: Tuesday, January 6, 2026 8:34 PM To: MarksLab-DasLab/RNAGym ***@***.***> Cc: YuanLi-X ***@***.***>; Author ***@***.***> Subject: Re: [MarksLab-DasLab/RNAGym] Being confused by the datastructure (Issue #15) [https://avatars.githubusercontent.com/u/899362?s=20&v=4]murfalo left a comment (MarksLab-DasLab/RNAGym#15)<#15 (comment)> Hello! Thank you for your feedback. I believe this may be a duplicate of another issue<#7 (comment)>. In short, the difference between HuggingFace and our self-hosted files is indeed a bug, the current documentation is confusing, and we are actively working to address this. We plan to address this with clarified steps for reproduction in the next edition of RNAGym which will be released in the coming months. Thank you for your patience. — Reply to this email directly, view it on GitHub<#15 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BOFZRIVLZT2KWVA6WQM2NVT4FQL43AVCNFSM6AAAAACQ3GDC4WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTOMJWGIZDSNBXGE>. You are receiving this because you authored the thread.Message ID: ***@***.***>