claudeVIDEO
Opus 5.5 makes an exaggerated explainer video on PPO, GRPO and DPO
The author used Opus 5.5 to make an over-the-top video explaining the PPO, GRPO and DPO reinforcement learning alignment algorithms, and says it turned out better than expected.
Dongxi 东锡 NLP
@dongxi_nlpArchived 2026.09.28Originally posted 2026.09.28
Dongxi 东锡 NLP @dongxi_nlpOriginally posted 2026.09.28
我觉得 Opus 5.5 不抽大麻，做不出这种东西。 看完你就懂 PPO, GRPO, DPO! Better than expected.
x.com/dongxi_nlp/status/2104350237062021225Archived 2026.09.28 10:10