Shivaram Venkataraman

Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs

PLoRA: Efficient LoRA Hyperparameter Tuning for Large Language Models

PipeTune: Tuning Pipeline Parallelism for Efficient Vision-Language Model Training

What Limits Agentic Systems Efficiency?

Scaling Inference-Efficient Language Models

LV-XAttn: Distributed Cross-Attention for Long Visual Inputs in Multimodal Large Language Models