Vatsal Baherwani — accepted AI-conference papers

2026

DynaGuard: A Dynamic Guardian Model With User-Defined Policies

Monte Hoover, Vatsal Baherwani, Neel Jain, Khalid Saifullah, Joseph James Vincent, Chirag Jain +4

ICLR 2026poster

Guardian models play a crucial role in ensuring the safety and ethical behavior of user-facing AI applications by enforcing guardrails and detecting harmful content. While standard guardian models are limited to predefined, static harm categories, we introduce DynaGuard, a suite of dynamic guardian…

Cited by 0 Source ↗code ↗Scholar ↗

2025

Dense Backpropagation Improves Training for Sparse Mixture-of-Experts

Ashwinee Panda, Vatsal Baherwani, Zain Sarwar, Benjamin Thérien, Sambit Sahu, Tom Goldstein +1

NeurIPS 2025poster

Mixture of Experts (MoE) pretraining is more scalable than dense Transformer pretraining, because MoEs learn to route inputs to a sparse set of their feedforward parameters. However, this means that MoEs only receive a sparse backward update, leading to training instability and suboptimal performanc…

Cited by 0 Source ↗code ↗Scholar ↗