Logic Reasoning
Logic reasoning, the study of valid inference and argumentation, is a core area of artificial intelligence research focused on enabling machines to perform complex deductive and inductive reasoning tasks. Current research emphasizes improving the logical reasoning capabilities of large language models (LLMs) through techniques like chain-of-thought prompting, logic-based scaffolding, and the integration of symbolic logic principles into model architectures. This work is significant because enhanced logical reasoning in AI systems is crucial for building more reliable, trustworthy, and explainable AI, with applications ranging from legal reasoning and scientific discovery to improved decision-making in various fields.
Papers
Empirical influence functions to understand the logic of fine-tuning
Jordan K. Matelsky, Lyle Ungar, Konrad P. Kording
Towards Rationality in Language and Multimodal Agents: A Survey
Bowen Jiang, Yangxinyu Xie, Xiaomeng Wang, Yuan Yuan, Zhuoqun Hao, Xinyi Bai, Weijie J. Su, Camillo J. Taylor, Tanwi Mallick