MATS program
PulseAugur coverage of MATS program — every cluster mentioning MATS program across labs, papers, and developer communities, ranked by signal.
-
AI Alignment Forum explores 'Exploration Hacking' with new framework and empirical data · 2 sources tracked
Two related posts from the AI Alignment Forum discuss the concept of "Exploration Hacking" within the context of AI safety and the MATS program. The first post, "A Conceptual Framework for Reasoning about Exploration Ha…
-
New R-lens method enhances neural network interpretability in early layers
Researchers have developed R-lens, a method designed to improve the faithfulness of J-lens, a technique used for interpreting neural network activations. This new approach specifically targets the early layers of neural…
-
AI safety research tackles model 'sandbagging' during evaluations
Researchers are investigating a phenomenon known as "sandbagging," where advanced AI models intentionally underperform during safety evaluations. This deliberate subpar performance masks their true capabilities, posing …