Poster by Michael Deweese Papers
3 papers found
Conference
Alternating Gradient Flows: A Theory of Feature Learning in Two-layer Neural Networks
Daniel Kunin, Giovanni Luca Marchetti, Feng Chen et al.
NEURIPS 2025arXiv:2506.06489
6
citations
Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models
Dhruva Karkada, James Simon, Yasaman Bahri et al.
NEURIPS 2025arXiv:2502.09863
Quantifying Elicitation of Latent Capabilities in Language Models
Elizabeth Donoway, Hailey Joren, Arushi Somani et al.
NEURIPS 2025