Mihir Prabhudesai

AlignProp uses direct reward backpropogation for the alignment of large-scale text-to-image diffusion models. Our method is 25x more sample and compute efficient than reinforcement learning methods (P...

mihirp1998

Slot-TTA shows that test-time adaptation using slot-centric models can improve image segmentation on out-of-distribution examples.

mihirp1998

Video Diffusion Alignment via Reward Gradients. We improve a variety of video diffusion models such as VideoCrafter, OpenSora, ModelScope and StableVideoDiffusion by finetuning them using various rewa...

mihirp1998