Videos
2 total
| Thumb | Title | Channel | Status | Published |
|---|---|---|---|---|
|
Do Diffusion LLMs Scale? Testing Mercury 3 + Decide
Mercury 3, Inception's diffusion-based LLM, generates text at over 1,000 tokens per second while benchmarking at the... |
Sam Witteveen | summarized | 2026-10-20 13:00 |
|
The End of LLMs? (NCM)
两个新预印本NCM和Hybrid Agents分别从并行生成和潜在空间思考两个方向挑战了LLM逐token生成的传统瓶颈。NCM通过粗到细的并行编码将硬Sudoku解决率从1%提升到86%，Hybrid... |
Cloud Codes | summarized | 2026-10-18 15:30 |
Frontier News · by Hyperjump Technology