AI RESEARCH
Dual-objective Language Models: Training Efficiency Without Overfitting
arXiv CS.AI
•
ArXi:2512.14549v3 Announce Type: replace-cross This paper combines autoregressive and masked-diffusion