Go to ACL ARR 2025 February homepageLogical DA: Enhancing Data Augmentation for Logical Reasoning via a Multi-Agent System
Recent advancements in large language models (LLMs) have highlighted the importance of improving their reasoning capabilities. A critical challenge lies in the scarcity of high-quality reasoning data—characterized by diversity and rich supervisory signals—necessary for robust model training. While data augmentation (DA) methods have been leveraged to mitigate this scarcity, prevailing approaches often introduce noise and exhibit logical inconsistencies, thereby diminishing their utility for complex reasoning tasks. Moreover, existing DA paradigms predominantly isolate data synthesis from label validation, failing to unify these complementary processes within a cohesive architecture. % A major challenge, however, is the scarcity of high-quality reasoning data, which is diverse and rich in supervisory signals. While data augmentation (DA) techniques have been employed to address this issue, existing methods often produce noisy data and lack logical consistency, limiting their effectiveness in reasoning tasks. Moreover, current DA approaches typically focus on either data generation or label verification, without integrating both tasks in a unified framework. To address these limitations, we introduce Logical DA, a multi-agent framework for enhancing reasoning-focused data augmentation in few-shot learning scenarios. % In this paper, we propose Logical DA, a multi-agent system designed to enhance reasoning data augmentation in few-shot learning scenarios. Our system includes four agents operating through two synergistic phases: (1) diverse data generation, and (2) label verification. % , which together ensure data diversity and label consistency through two main stages: (1) diverse data generation, and (2) label verification. The system incorporates a reflection mechanism to continuously improve data quality by leveraging feedback from logical validation. We demonstrate the effectiveness of Logical DA through experiments on various tasks and datasets, achieving the highest average improvement in task accuracy in both fine-tuning and in-context learning paradigms, with an average improvement of 7.61% when applied to fine-tuning.