All Publications
Masked Generative Nested Transformers with Decode Time Scaling
Sahil Goyal, Debapriya Tula, Pradeep Shenoy, Prateek Jain, Sujoy Paul
ICML 2025VisionEfficiencyConditional Computation
TL;DR: An efficient framework for progressive decoding with nested models for faster inference.
Bayesian Collaborative Bandits with Thompson Sampling for Improved Outreach in Maternal Health
Arpan Dasgupta, Arun Suggala, Karthikeyan Shanmugam, Milind Tambe, Aparna Taneja
AAMAS 2025Reinforcement LearningTheorySocietal Impact
TL;DR: An improved collaborative bandits approach with bayesian regret derivation for a special case!
Mixture of Nested Experts: Adaptive Processing of Visual Tokens
Nidhi Hegde, Aditya Kusupati, Arsha Nagrani, Shyamal Buch, Prateek Jain, Anurag Arnab, Sujoy Paul
NeurIPS 2024Mixture of ExpertsConditional Computation
TL;DR: Token-wise routing between nested experts for tackling redundancy in visual modalities.
LookupViT: Compressing Visual Information to a smaller number of tokens
Rajat Koner, Gagan Jain, Prateek Jain, Volker Tresp, Sujoy Paul
ECCV 2024VisionEfficiency
TL;DR: An asyncronous version of attention with sub-quadratic scaling and superior performance.