ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.
Qwq-32b: Embracing the power of reinforcement learning, March 2025
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
ReTool: Reinforcement Learning for Strategic Tool Use in LLMs
ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.