AI RESEARCH
Universal Reasoner: A Single, Composable Plug-and-Play Reasoner for Frozen LLMs
arXiv CS.LG
•
ArXi:2505.19075v3 Announce Type: replace-cross Large Language Models (LLMs) have nstrated remarkable general capabilities, but enhancing skills such as reasoning often demands substantial computational resources and may compromise generalization. While Parameter-Efficient Fine-Tuning (PEFT) methods offer a resource-conscious alternative, they typically require re