A new method from Amazon scientists, presented at the International Conference on Machine Learning (ICML), improves LLM-as-a-judge evaluation by accounting for correlations between judges' outputs, ou…
MIT CSAIL and Tsinghua University researchers developed GeoPT, an AI pre-training method that learns physics from 1.3 million synthetic particle interactions, cutting labeled data needs by 60% and ach…
Researchers at the International Conference on Machine Learning presented evidence that a fundamental architectural flaw in large language models makes them inherently impossible to fully secure again…
Amazon researchers presented ControlG, a framework for multi-objective graph self-supervised learning, at ICML 2026. ControlG uses a proportional-integral-derivative (PID) controller from industrial c…
Shinsegae Department Store's AI-powered hyperpersonalization technology, developed jointly with Seoul National University, has been accepted for presentation at the International Conference on Machine…
MIT CSAIL and Toyota Research Institute have developed SceneSmith, a framework using three AI agents powered by GPT-5.2 to build detailed 3D virtual environments for robot training, generating scenes …
MIT researchers, led by graduate student Vinith Suriyakumar and associate professors Ashia Wilson and Marzyeh Ghassemi, in collaboration with child safety nonprofit Thorn, have developed a new auditin…
IBM researchers unveiled a new generative AI model architecture called Continued Fraction Generative Networks (CoFrGeNets) at the International Conference on Machine Learning in Seoul, which fundament…
Researchers found that smaller language models suffer from 'embedding condensation,' where token embeddings collapse into a narrow cone, reducing expressivity. They introduced a dispersion loss to cou…