Raghav Singhal

Safety Subspaces are Not Linearly Distinct: A Fine-Tuning Case Study

ABBA-Adapters: Efficient and Expressive Fine-Tuning of Foundation Models

Initialization using Update Approximation is a Silver Bullet for Extremely Efficient Low-Rank Fine-Tuning

Fed-SB: A Silver Bullet for Extreme Communication Efficiency and Performance in (Private) Federated LoRA Fine-Tuning

FedEx-LoRA: Exact Aggregation for Federated and Efficient Fine-Tuning of Foundation Models

M3CoL: Harnessing Shared Relations via Multimodal Mixup Contrastive Learning for Multimodal Classification