On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning

On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning explores This research identifies a critical flaw in how large language models learn reasoning from diverse data sources and proposes a filtering method to significantly improve generalization performance on complex reasoning tasks.. Commercial viability score: 7/10 in LLM Reasoning.