Power law graph attention: exact generalization of scaled dot-product attention, empirical collapse at inference Paper • 2608.10288 • Published 5 days ago • 7
On-Policy Delta Distillation for Multilingual Math Reasoning Paper • 2608.05802 • Published 9 days ago • 32
AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling Paper • 2608.02602 • Published 12 days ago • 79