How to Detect and Defeat Molecular Mirage: A Metric-Driven Benchmark for Hallucination in LLM-based Molecular Comprehension

TL;DR

This paper introduces Mol-Hallu, a new metric for evaluating hallucination in LLMs for molecular understanding, and proposes HRPP to reduce hallucinations, enhancing model reliability in scientific tasks.

Contribution

It presents Mol-Hallu, a novel metric for assessing hallucination in molecular LLMs, and proposes HRPP, a post-processing method to mitigate hallucinations in these models.

Findings

Mol-Hallu effectively quantifies hallucination levels in molecular LLMs.

HRPP significantly reduces hallucination in both decoder-only and encoder-decoder models.

Analysis reveals knowledge shortcut phenomenon as a key source of hallucination.

Abstract

Large language models are increasingly used in scientific domains, especially for molecular understanding and analysis. However, existing models are affected by hallucination issues, resulting in errors in drug design and utilization. In this paper, we first analyze the sources of hallucination in LLMs for molecular comprehension tasks, specifically the knowledge shortcut phenomenon observed in the PubChem dataset. To evaluate hallucination in molecular comprehension tasks with computational efficiency, we introduce \textbf{Mol-Hallu}, a novel free-form evaluation metric that quantifies the degree of hallucination based on the scientific entailment relationship between generated text and actual molecular properties. Utilizing the Mol-Hallu metric, we reassess and analyze the extent of hallucination in various LLMs performing molecular comprehension tasks. Furthermore, the Hallucination Reduction Post-processing stage~(HRPP) is proposed to alleviate molecular hallucinations, Experiments show the effectiveness of HRPP on decoder-only and encoder-decoder molecular LLMs. Our findings provide critical insights into mitigating hallucination and improving the reliability of LLMs in scientific applications.