Uchicago DSI - Why Can’t Powerful LLMs Learn Multiplication?
Oct 24, 2025 - This paper’s key contribution is architectural insights and training techniques can overcome obstacles that scaling alone cannot address.
How to build AI scaling laws for efficient LLM training and budget maximization
Sep 16, 2025 - MIT and MIT-IBM Watson AI Lab researchers have developed a universal guide for estimating how large language models (LLMs) will perform based on smaller models in the same family.
Unpacking the bias of large language models
Jun 17, 2025 - MIT researchers discovered the underlying cause of position bias, a phenomenon that causes large language models to overemphasize the beginning or end of a document or conversation, while neglecting the middle.
Could LLMs help design our next medicines and materials?
Apr. 9, 2025 - A new multimodal tool combines a LLM with powerful graph-based AI models to efficiently find new, synthesizable molecules with desired properties, based on a user’s queries in plain language.
MIT, Cornell Research AI Chatbots to Combat Conspiracy Theories
Researchers from MIT Sloan and Cornell University found that chatting about a conspiracy theory with a large language model reduced people’s belief in it by about 20 percent.