Release the medical dataset generated by KISA

[IcecreamArtist]

Hello,

I'm very interested in your excellent work on MedResearcher-R1. Your paper describes a medical trajectory dataset generated through KISA from 30M+ PubMed abstracts, which appears to be a key contribution to achieving the impressive results on MedBrowseComp.

I noticed you've also open-sourced a high-quality QA dataset within your GitHub repository (specifically TrajectoryGenerationPipeline/qa_data/open_data.jsonl). However, upon examination, this dataset appears to focus on general-domain questions rather than medical-specific content, and may not be the actual dataset used for training MedResearcher-R1.

If you're interested, could you please release the dataset used in training MedResearcher-R1? Let me know if I have any misunderstanding.

Thank you very much!

Sincerely,
IcecreamArtist

[cz3716087]

Hello,

I'm very interested in your excellent work on MedResearcher-R1. Your paper describes a medical trajectory dataset generated through KISA from 30M+ PubMed abstracts, which appears to be a key contribution to achieving the impressive results on MedBrowseComp.

I noticed you've also open-sourced a high-quality QA dataset within your GitHub repository (specifically TrajectoryGenerationPipeline/qa_data/open_data.jsonl). However, upon examination, this dataset appears to focus on general-domain questions rather than medical-specific content, and may not be the actual dataset used for training MedResearcher-R1.

If you're interested, could you please release the dataset used in training MedResearcher-R1? Let me know if I have any misunderstanding.

Thank you very much!

Sincerely, IcecreamArtist

Hi @IcecreamArtist,

Thank you for your interest in our work and for your thoughtful comment!
We do have plans to release additional medical-domain datasets related to MedResearcher-R1. However, since parts of the data are connected to internal corpora and are subject to company privacy and security review, the release process requires further approval.

We appreciate your patience and understanding. Please stay tuned — any updates about dataset releases will be announced in this repository as soon as they become available.

Best regards,
cz3716087