AI RESEARCH
Why Do Reasoning Models Lose Coverage? The Role of Data and Forks in the Road
arXiv CS.LG
•
ArXi:2605.17026v1 Announce Type: new Recent progress in large language models has led to the emergence of reasoning models, which have shown strong performance on complex tasks through specialized fine-tuning procedures. While these methods reliably improve pass accuracy, prior works have observed that they show a coverage shrinkage behavior, where pass degrades relative to the base model. In this paper, we investigate the reasoning shrinkage arise under SFT-based post-