How-to-Make Smaller LLMs R1-Smart (UC Berkeley)
All right w/ authors:
"Climbing the Ladder of Reasoning: What LLMs Can—and
Still Can’t—Solve after SFT?"
Yiyou Sun1, Georgia Zhou1, Hao Wang1, Dacheng Li1, Nouha Dziri2, Dawn Song1
1 University of California, Berkeley,
2 Allen Institute for AI
arXiv:2504.11741v1
@UCBerkeley