29. SNR adaptive signSGD view of Adam is good. by @YouJiacheng (You Jiacheng) · backlist 2026-05-06 · rubric 95.0
88. 1% refers to the cache hit price, no problem there. But what he said is indeed wrong; the agent's high cache hit … by @YouJiacheng (You Jiacheng) · backlist 2026-05-06 · rubric 84.0
82. larger compute prefer smaller vocabulary, interesting. (t.co) by @YouJiacheng (You Jiacheng) · backlist 2026-05-05 · rubric 82.0