from Hacker News

Understanding R1-Zero-Like Training: A Critical Perspective [pdf]

by delifue on 3/22/25, 7:59 AM with 1 comments

• by delifue on 3/22/25, 8:01 AM

  From: https://x.com/zzlccc/status/1903162768083259703

  DeepSeek-V3-Base already exhibits "Aha moment" before RL-tuning

  The ever-increasing output length in RL-tuning might be due to a BIAS in GRPO