Problem-solving data in continued pretraining improves LLM math reasoning more than general math corpora, and tutorship amplification is the most effective synthesis method.
Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages
Problem-solving data in continued pretraining improves LLM math reasoning more than general math corpora, and tutorship amplification is the most effective synthesis method.