Author

Soumik Mukhopadhyay

h-index 1 View on Semantic Scholar

We have 1 of 6 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Jul 2026

Selective Timestep Weighting and Advantage-Based Replay for Sample-Efficient Diffusion RLHF

This paper introduces a per-timestep weighting scheme that reweights denoising steps during policy optimization and introduces a replay mechanism that prioritizes informative trajectories, enabling the model to reuse past samples instead of repeatedly querying new rewards.

Eric Zhu, Abhinav Shrivastava, Soumik Mukhopadhyay · arXiv.org · 0 citations