Can we automatically evaluate the semantic accuracy of LLM translations without human annotation?
Our #ICLR ’25 work introduces a novel approach for assessment of truth maintenance in formal language translation.
Joint w/ Rushang Karia, Daksh Dobhal, @sidsrivast.bsky.social.
(1/3)

Can we align an AI system with users’ expectations when it has limited, noisy information about the real world?
Our #NeurIPS ’24 work allows users fuse high-level objectives with preferences and constraints based on the agent’s current belief about its environment.
Joint w/ @sidsrivast.bsky.social