Artificial intelligence is now widely applied in drug discovery and development, accelerating the entry of novel pharmaceuticals into clinical practice. As the variety of drugs expands, the workload for pharmaceutical inspection has increased significantly, making the demand for automated verification of pharmaceutical inspection results increasingly urgent. However, current research lacks reliable evaluation methods and datasets. To address this issue, we construct a Pharmaceutical Certificate of Analysis Dataset (PCoAD) based on large language model validation and correction. This dataset comprises 4,272 manually verified pharmaceutical certificates of analysis (PCoAs) based on Chinese and U.S. pharmacopoeias, comprehensively testing the ability of model to verify pharmaceutical inspection processes. Based on PCoAD, we propose the Multi-Agent Cooperation based on Adaptive Retrieval (MACAR) framework. This framework employs text chunking, adaptive retrieval, and inference to validate PCoAs. Experimental results demonstrate that MACAR outperforms multiple state-of-the-art methods across various types of tasks.

Large Language Models (LLMs) frequently face limitations, such as judgment biases and cognitive deadlocks, in complex reasoning due to “Degradation of Thought” in self-reflection. While Multi-Agent Debate (MAD) approaches attempt to address this issue, their inherently adversarial nature suppresses diverse perspectives by overemphasizing antagonism, which results in the loss of partially valid reasoning. In this paper, we propose Multi-Agent Diverse Thinking (MADT), a novel cognitive framework that redefines multi-agent collaboration by shifting from adversarial confrontation toward constructive synergy. To break the oppositional deadlock found in traditional debates, MADT decomposes the thinking process into fine-grained modules: an Affirmative Thinker to preserve rational components, a Critical Thinker to rectify errors, and a Growth-Minded Thinker to provide optimization suggestions. These agents jointly foster a synergistic environment where specialized roles protect valid logic while iteratively refining flaws. Meanwhile, a Leader agent coordinates the collaborations and synthesizes the feedback from multiple thinkers to produce the final result. Extensive experiments on Common Machine Translation and Counter-Intuitive Arithmetic Reasoning tasks show that MADT consistently outperforms Self-Reflection and MAD baselines, validating the superior ability to enhance complex logical and reasoning performance in LLMs.