Tensor Times
An academic-grade journal of artificial intelligence research — peer-reviewed analysis of large language models, machine learning frameworks, and the engineering behind today’s most consequential AI systems.
Latest Analysis
-
First 48 Hours of Harbor: 11 Labs Already Submitting Results, Leaderboard Live
In just 48 hours, DeepMind’s Harbor open-source harness has seen 11 labs submit results, with a leaderboard showing significant performance gaps.
-
University of Hawaiʻi Algorithm Advances Physics-Informed Machine Learning
A new algorithm from the University of Hawaiʻi significantly enhances physics-informed machine learning by enforcing conservation laws as hard constraints.
-
MIT Team Applies Control Theory to Optimize Large Language Model Training
MIT researchers have developed a method using control theory to reduce compute costs in training large language models by up to 35% without quality loss.
-
Cornell Neuro-Symbolic AI Architecture Achieves 100x Energy Efficiency Gains with Improved Accuracy
A Cornell University research team has demonstrated a neuro-symbolic AI system that reduces energy consumption by up to 100 times while outperforming transformer baselines on reasoning benchmarks, pot
-
EU Commission Proposes Mandatory ‘Ethical Impact Statements’ for High-Risk AI Systems
The European Commission unveiled a draft regulation this morning requiring developers of high-risk AI systems to submit comprehensive ethical impact statements before deployment, marking the most sign
-
The Cognitive Architecture of In-Context Learning: Neuroscientific Perspectives on Few-Shot Adaptation Mechanisms in Transformer Models
Introduction: Bridging Artificial and Biological Intelligence The remarkable ability of large language models (LLMs) to perform new tasks from just a few examples, a phenomenon known as in-context…