-
Two ways for a model to say "I don't know"
Kendall and Gal (NeurIPS 2017) give the standard recipe for modeling aleatoric and epistemic uncertainty jointly in a single network, on segmentation and depth regression.
-
Learning without a gradient step
Xie, Raghunathan, Liang, and Ma (ICLR 2022) give a formal Bayesian account of why in-context learning improves with more demonstrations without any gradient update.
-
A task, folded into a single vector
Todd, Li, Sharma, Mueller, Wallace, and Bau (ICLR 2024) show that in-context learning induces a single, portable vector that causally encodes the task.
-
Google Gemini updates: Flash 1.5, Gemma 2 and Project Astra
We’re sharing updates across our Gemini family of models and a glimpse of Project Astra, our vision for the future of AI assistants.
-
Displaying External Posts on Your al-folio Blog