LLMs as Noisy Channels: A Shannon Perspective on Model Capacity and Scaling Laws Paper • 2605.23901 • Published 12 days ago • 13
Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information Paper • 2605.11609 • Published 22 days ago • 195
Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention Paper • 2605.22791 • Published 13 days ago • 30
Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation Paper • 2604.10098 • Published Apr 11 • 82