一作ACL 2026Poster· Natural Language Processing
Student Guides Teacher: Weak-to-Strong Inference via Spectral Orthogonal Exploration
Dayu Wang · Jiaye Yang · Weikang Li · Jiahui Liang · Yang Li · Deguo Xia · Jizhou Huang
摘要
Large Language Models (LLMs) often suffer from “Reasoning Collapse” on challenging mathematical reasoning tasks, where stochastic sampling produces lexical variations of the same erroneous logic rather than genuine semantic exploration. We observe that failed reasoning traces are often associated...