Research

Naive idea for diffusion guidance with non-differentiable guidance function

Background on discrete diffusion and a proposed diffusion method for inverse protein folding

Proposed method of quantifying groundedness of LLM output through data augmentations

Proposed method of quantifying uncertainty of LLMs

A research proposal to improve offline learning using novelty driven policy distillation