大模型推理能力

MedRule-KG: A Knowledge-Graph--Steered Scaffold for Mathematical Reasoning with a Lightweight Verifier

MedRule-KG：一种由知识图谱引导的轻量级验证器支持的数学推理框架

Crystal Su

专题命中 数学推理 ：reasoning(title,abstract)；verifier(title,abstract)；分类 cs.AI

AI总结 MedRule-KG通过结合知识图谱和轻量级验证器，提升数学推理的准确性和规则一致性。

Comments This paper is withdrawn due to issues with attribution and citation accuracy

MedRule-KG: A Knowledge-Graph--Steered Scaffold for Reliable Mathematical and Biomedical Reasoning

MedRule-KG: 一种基于知识图谱的可靠数学和生物医学推理框架

Crystal Su

专题命中 数学推理 ：reasoning(title,abstract)；chain-of-thought(abstract)；verifier(abstract)；分类 cs.AI

AI总结 MedRule-KG通过知识图谱框架和轻量验证器，提升科学推理和药物发现的可靠性与准确性。

Comments This paper is withdrawn due to issues with attribution and citation accuracy

From Next-Token to Mathematics: The Learning Dynamics of Mathematical Reasoning in Language Models

从单个词到数学：语言模型中数学推理学习动态

Shubhra Mishra, Gabriel Poesia, Noah D. Goodman

机构 * Department of Computer Science 1 and Psychology 2（计算机科学系和心理学系） ； Stanford University（斯坦福大学） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.CL、cs.AI

AI总结 本研究分析了语言模型在预训练和后训练过程中数学推理能力的发展，揭示了数学技能学习顺序与课程设计的相关性，并探讨了指令微调对不同数学技能的影响。

Comments Accepted to COLM 2025. Dataset and code: https://github.com/gpoesia/mathcamps/

Journal ref Conference on Language Modeling (COLM), Montreal, Canada, October 7-10, 2025

TRACE: A Framework for Analyzing and Enhancing Stepwise Reasoning in Vision-Language Models

TRACE：一种用于分析和增强视觉-语言模型逐步推理的框架

Shima Imani, Seungwhan Moon, Lambert Mathias, Lu Zhang, Babak Damavandi

机构 * Meta Reality Lab（Meta现实实验室） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.AI

AI总结 TRACE通过辅助推理集和一致性评估，提升视觉-语言模型的推理透明性和错误检测能力，促进模型改进和优化。