Scaling Properties of Same-Family On-Policy Distillation Paper • 2609.32722 • Published 5 days ago • 203
When Does Dense Retrieval Need Asymmetric Geometry? A Bias-Variance Theory of Shared and Dual Projections Paper • 2609.32488 • Published 5 days ago • 13
Selecting The Most Informative Tokens in Natural Language Autoencoders Paper • 2609.37040 • Published 2 days ago • 12
ROSS: Relearning from Self-Generated Rollouts through Selective Supervision Paper • 2609.35954 • Published 3 days ago • 33