Plan to release the SFT dataset

[Yinhuaihong]

This is very impressive work! I was wondering if there are any plans to release the SFT dataset? Doing so would greatly enhance the reproducibility of your work and facilitate broader adoption of the benchmark within the community.

[bcol23]

Thank you for your interest and kind words!

Regarding the SFT dataset, unfortunately, we are unable to release it due to data licensing restrictions. Specifically, the agreement prevents us from distributing the results returned by the tool APIs.

However, we have found that performing RL directly on the Instruct/Chat models also yields very competitive results. You can refer to the following example scripts in the repository to try this approach:

• scripts/deepresearch/run-qwen3-8B.sh
• scripts/travel/run-qwen3-8B.sh

Hope this helps!