Logically Consistent Language Models via Neuro-Symbolic Integration

TL;DR

This paper introduces a neuro-symbolic loss function that enhances large language models' logical consistency and ability to reason reliably by integrating external facts and rules, even with limited fine-tuning.

Contribution

It presents a novel neuro-symbolic training approach that improves LLMs' logical consistency, enables combining multiple constraints, and enhances extrapolation to unseen factual data.

Findings

Improves self-consistency of LLMs with limited fine-tuning.

Allows combining multiple logical constraints effectively.

Enhances extrapolation to unseen but related factual knowledge.

Abstract

Large language models (LLMs) are a promising venue for natural language understanding and generation. However, current LLMs are far from reliable: they are prone to generating non-factual information and, more crucially, to contradicting themselves when prompted to reason about relations between entities of the world. These problems are currently addressed with large scale fine-tuning or by delegating reasoning to external tools. In this work, we strive for a middle ground and introduce a loss based on neuro-symbolic reasoning that teaches an LLM to be logically consistent with an external set of facts and rules and improves self-consistency even when the LLM is fine-tuned on a limited set of facts. Our approach also allows to easily combine multiple logical constraints at once in a principled way, delivering LLMs that are more consistent w.r.t. all constraints and improve over several baselines w.r.t. a given constraint. Moreover, our method allows LLMs to extrapolate to unseen but semantically similar factual knowledge, represented in unseen datasets, more systematically.