Author

Lavanya Nigam

View on Semantic Scholar

1 paper indexed here

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.

Preprint Aug 2026

Lost in Interpolation: Why Predictive Feedback Fails in Diffusion Language Models

Spherical Soft-Masking (S-SM), a drop-in replacement that aggregates the top-(k) predictions with a Fr'echet mean on the hypersphere and blends this mean with the mask direction using spherical linear interpolation (SLERP), then restores the native mask norm.

Lavanya Nigam, Ishaan Bansal, Aryan Sood et al. · 0 citations