Lav R. Varshney

A unified Pāninian framework could revolutionize NLP for over a billion speakers by merging disparate language resources into a single, high-performance system.

You can now audit Rényi differential privacy with near-optimal sample complexity, thanks to a new framework that directly estimates Rényi divergence using Donsker-Varadhan estimators.

LLM hallucinations aren't random errors, but predictable consequences of overlapping "hallucination basins" in the model's latent space, offering a new path to control them.