https://blog.ceshine.net/tags/llm/ Recent content in Llm on Veritable Tech Blog Hugo -- 0.149.1 en 2015-2026 Ceshine Lee · CC BY-NC-SA 4.0 Sat, 11 Apr 2026 00:00:00 +0000 https://blog.ceshine.net/post/PPO-GRPO-GSPO/ Sat, 11 Apr 2026 00:00:00 +0000 https://blog.ceshine.net/post/PPO-GRPO-GSPO/ A walkthrough of PPO, GRPO, and GSPO — covering clipped surrogates, group-relative advantages, and sequence-level importance sampling.