Results 311 - 320 of 24331
First Proof has released the results of its second batch benchmark, assessing the ability of AI systems to autonomously solve naturally occurring mathematical research problems. Their solutions were evaluated by thirty expert mathematicians in a gathering last week at Harvard’s Center for Mathematical Sciences and Applications.
No abstract available.
This talk briefly describes three projects. The first two focus on evaluating world modeling capabilities in frontier models through the evaluation of dual perspective reasoning and knot manipulation. The third part describes the state separation hypothesis, which posits that mechanically separating prediction from state computation in LLMs benefits performance.
Objects play an important role in vision. Much of human vision is centered around objects and there is evidence we develop a basic understanding of what objects are from a very young age.
Learning about objects without supervision has been a focus of much research in recent years. Many models have been suggested with different structures, assumptions and inductive biases
and while impressive progress has been achieved in some limited domains we have yet to obtain a general system that can learn about objects unsupervised from real-world data.
In this talk I argue that there is a fundamental mismatch between some of the assumptions made by most "object centric" models and real-world data, and that this mismatch prevents such models from learning
meaningful representations at scale - ultimately making the problem setting ill-posed. Following that I will present some of our current work which attempts to address some of these issues.
In current large language models, learning is based on supervised training using gigantic language datasets. Human learning begins differently: before having significant language capabilities, infants acquire a broad range of meaningful concepts, relationships between concepts and their implications, with little or no supervision.
I will describe examples of modeling infant-like acquisition of meaningful concepts and early conceptual structures. I will use the results to make comparisons with AI models, and discuss whether early conceptual structures can contribute to them.