Investigating the (De)Composition Capabilities of Large Language Models in Natural-to-Formal Language Conversion

TL;DR

This paper assesses the fundamental decomposition and composition abilities of large language models in converting natural language to formal language, revealing significant deficiencies and error types that impact their robustness and generalization.

Contribution

It introduces the DEDC framework for semi-automatic evaluation of LLMs' decomposition and composition capabilities in N2F, providing a new perspective for analysis and improvement.

Findings

LLMs are deficient in both decomposition and composition.

Error types relate to natural language understanding and symbolic system use.

Both compositional gaps and counter-intuitive symbolic names hinder LLM performance.

Abstract

To achieve generalized and robust natural-to-formal language conversion (N2F), large language models (LLMs) need to have strong capabilities of decomposition and composition in N2F when faced with an unfamiliar formal language and be able to cope with compositional gaps and counter-intuitive symbolic names. To investigate whether LLMs have this set of basic capabilities in N2F, we propose the DEDC framework. This framework semi-automatically performs sample and task construction, allowing decoupled evaluation of the set of decomposition and composition capabilities of LLMs in N2F. Based on this framework, we evaluate and analyze the most advanced LLMs, and the main findings include that: (1) the LLMs are deficient in both decomposition and composition; (2) the LLMs show a wide coverage of error types that can be attributed to deficiencies in natural language understanding and the learning and use of symbolic systems; (3) compositional gaps and counter-intuitive symbolic names both affect the decomposition and composition of the LLMs. Our work provides a new perspective for investigating the basic capabilities of decomposition and composition of LLMs in N2F. The detailed analysis of deficiencies and attributions can help subsequent improvements of LLMs.