This work was co-led by Ali Shirali and me, along with @_ahmedmalaa. For more information see: https://t.co/VBx7nZIWxk arXiv:

We evaluate our method in off-policy and offline settings using simulated environments and real health records from intensive care units, showing strong performance compared to common offline RL techniques.

RL approaches in healthcare often only reward models based on sparse outcomes such as 30-day mortality. Our solution? Prune the action space before learning. By doing so, we can effectively integrate intermediate biomarker signals without distorting the main objective.

How can reinforcement learning better leverage ICU biomarker signals for decision-making? Our new paper introduces PrunedQL—a two-stage algorithm that identifies actions to avoid using these signals, then optimizes patient outcomes. Check it out! https://t.co/VBx7nZIWxk

Excited to share the camera ready of our #ICLR2024 paper "InstructCV: Instruction-Tuned Text-to-Image Diffusion Models as Vision Generalists" led by @yule_gan, Sungwoo Park and @alexmschubert Paper: https://t.co/8v0BjNIVla Code: https://t.co/m4rKpwo6G9

Excited to share a new preprint! InstructCV is a language interface for vision tasks trained by instruction-tuning a text-to-image generative model #StableDiffusion 🧵 📌 Paper: https://t.co/8v0BjNJtaI 📌 Code: https://t.co/m4rKpwoEvH 📌 @Gradio demo: https://t.co/pLKikKj26E

Generative models can synthesize medical images but assessing quality & clinical validity is challenging. In this work w/ @Harry_S_Sun @GGoldgof @atulbutte, we use human feedback for evaluation & finetuning... 🧵 (1/4) Paper: https://t.co/4JUQoBcn8z Code: https://t.co/C2uI4XdxAF