ChemAU: Harness the Reasoning of LLMs in Chemical Research with Adaptive Uncertainty Estimation

2025-06-01

AI Summary

To address hallucination and knowledge gaps in large language models (LLMs) for chemical reasoning, this paper proposes an adaptive uncertainty estimation framework. It dynamically assigns uncertainty weights per reasoning step to enable fine-grained error localization, and tightly couples molecular representations and reaction rules—domain-specific expert modules—to inject targeted knowledge at weak reasoning steps, thereby correcting flawed reasoning chains. Unlike generic uncertainty estimation methods, which fail to pinpoint critical knowledge deficiencies, our approach introduces the first “position-aware, knowledge-enhanced” co-design mechanism. Evaluated on three chemical reasoning benchmarks, it significantly improves reasoning accuracy across three mainstream LLMs (average +18.7%) and substantially enhances uncertainty calibration (expected calibration error reduced by 42.3%), demonstrating both effectiveness and strong cross-model generalizability.

Technology Category

• Reasoning under Uncertainty: Uncertainty Representations
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) are widely used across various scenarios due to their exceptional reasoning capabilities and natural language understanding. While LLMs demonstrate strong performance in tasks involving mathematics and coding, their effectiveness diminishes significantly when applied to chemistry-related problems. Chemistry problems typically involve long and complex reasoning steps, which contain specific terminology, including specialized symbol systems and complex nomenclature conventions. These characteristics often cause general LLMs to experience hallucinations during the reasoning process due to their lack of specific knowledge. However, existing methods are struggling to effectively leverage chemical expertise and formulas. Moreover, current uncertainty estimation methods, designed to mitigate potential reasoning errors, are unable to precisely identify specific steps or key knowledge. In this work, we propose a novel framework called ChemAU, which incorporates our adaptive uncertainty estimation method that applies different uncertainty values based on the position of reasoning steps within the whole reasoning chain. Leveraging this method, ChemAU identifies gaps in chemistry knowledge and precisely supplements chemical expertise with the specialized domain model, thereby correcting and updating the previously flawed reasoning chain. Our experiments with three popular LLMs across three chemistry datasets demonstrate that ChemAU significantly enhances both reasoning accuracy and uncertainty estimation.

Problem

• LLMs struggle with chemistry problems due to complex terminology and reasoning
• Existing methods fail to effectively use chemical expertise and formulas
• Current uncertainty estimation cannot pinpoint reasoning errors in chemistry

Innovation

• Adaptive uncertainty estimation for reasoning steps
• Specialized domain model supplements chemistry knowledge
• Enhances reasoning accuracy and uncertainty estimation