Leena Mathur
1.1K posts
@lmathur_
PhD student @SCSatCMU. I study multimodal social intelligence in AI systems across embodiments. prev research @GoogleDeepMind, @RobustAI, @USC, @Caltech, @EPFL
Meet the 2026 cohort of KH scholars! These 87 new scholars make up the most global Knight-Hennessy Scholars cohort to date, and will pursue degrees in 45 graduate programs across all seven graduate schools at @Stanford: knight-hennessy.stanford.edu/news/knight-he… (1/2)
People talk, listen, watch, think, and collaborate at the same time, in real time. We've designed an AI that works with people the same way. We share our approach, early results, and a quick look at our model in action. thinkingmachines.ai/blog/interacti…
Today USC launched a pathbreaking university-wide initiative that will leverage AI to accelerate our societal impact in fields ranging from the health sciences and security to business and the arts. Thanks to the generosity and vision of Mark and Mary Stevens, whose names will now grace the USC Mark and Mary Stevens School of Computing and Artificial Intelligence, we will attract even more world-class AI talent to USC. Read more about the $200M gift establishing USC’s leadership in the Age of AI: uscne.ws/7e26aa @USCStevensAI @USCViterbi
We present Compute Optimal Tokenization! 🔡 Common in LLM scaling works stick to one tokenizer, sweeping data/model size. But what happens when we control the tokenizer’s compression rate (bytes/token)? Here we sweep tokenizers, params, and data across compute budgets: [1/N]
Recipes for teaching language models to handle long inputs don't work equally well across model families. We wanted to know why—is it the architecture, the training data, or both? 🧵
