Teaching LLMs to Parallelize Their Own Generations

Researchers from the Massachusetts Institute of Technology and Google have developed Pasta, a novel technique that enables large language models to parallelize their own text generation by producing semantically independent parts of responses simultaneously. The system teaches models to identify and annotate opportunities for asynchronous generation, enabling a purpose-built inference system to decode multiple sub-parts […]

Weiwen Tian joins DiCarlo’s CoCoSys team

Harvard University graduate student Weiwen Tian has been recruited to join DiCarlo’s CoCoSys team. Weiwen’s thesis work focuses on developing computational models of visually guided grasping behaviors in humans. Her work seeks an understanding of how the visual system infers 3D spatial information from 2D retinal inputs to guide grasping, and how these visual representations […]

Qualcomm Innovation Fellowship Awarded to Purdue ECE PhD Students for ML-Powered Tool to assist modern Chip design

Two CoCoSys PhD students from Purdue University have been named winners of the highly competitive Qualcomm Innovation Fellowship for their groundbreaking research aimed at enhancing energy efficiency in modern computing systems. Sujay Pandit and Aradhana Mohan Parvathy are students of Anand Raghunathan, Silicon Valley Professor of Electrical and Computer Engineering. The fellowship recognizes innovative research […]