Question about the released checkpoint

#5
by Connor1213 - opened

Great work! I am applying this model to some downstream tasks and it shows great potential to handle with visual grounding tasks compared with other medical mllm.And I noticed that the paper shows that Lingshu and Lingshu-RL have differences. If the released model whether the RL training one or not.Thanks for any provided information.

xww033 changed discussion status to closed
Lingshu: MLLMs for Unified Multimodal Medical Understanding and Reasoning org

Great work! I am applying this model to some downstream tasks and it shows great potential to handle with visual grounding tasks compared with other medical mllm.And I noticed that the paper shows that Lingshu and Lingshu-RL have differences. If the released model whether the RL training one or not.Thanks for any provided information.

the released model is Lingshu (not RL version). We do not release RL version, because it does not significantly improve over the SFT version. We will try to explore more about RL in the next version.

xww033 changed discussion status to open
Connor1213 changed discussion status to closed

Sign up or log in to comment