view article Article Welcome Inkling by Thinking Machines +2 burtenshaw, merve, pcuenq, ariG23498 • 13 days ago • 130
Over-Tokenized Transformer: Vocabulary is Generally Worth Scaling Paper • 2501.16975 • Published Jan 28, 2025 • 35
Rethinking Shrinkage Bias in LLM FP4 Pretraining: Geometric Origin, Systemic Impact, and UFP4 Recipe Paper • 2606.20381 • Published Jun 18 • 10
DFlash Collection Block Diffusion for Flash Speculative Decoding • 23 items • Updated 29 days ago • 151