Great article, thanks for the clear explanation.
A small notation suggestion: in the “A Recap on GRPO” section, using R_i for the reward score instead of r_i could help distinguish it from the importance ratio r_{i,t} and reduce confusion.
Zheng
AyongZheng
AI & ML interests
None yet
Recent Activity
commentedon an article 8 days ago
From GRPO to DAPO and GSPO: What, Why, and How authored a paper 4 months ago
Qianfan-OCR: A Unified End-to-End Model for Document Intelligence liked a model 4 months ago
baidu/Qianfan-OCROrganizations
None yet
commented on From GRPO to DAPO and GSPO: What, Why, and How 8 days ago
authored a
paper 4 months ago
liked a
model 4 months ago
upvoted a collection 4 months ago
upvoted a paper 4 months ago
请教LongBench-v2上评测时输出异常问题
9
#2 opened over 1 year ago
by AyongZheng
liked a
dataset over 1 year ago
liked a
dataset about 3 years ago
updated a
model over 3 years ago