Towards Logically Consistent Language Models via Probabilistic Reasoning
This addresses reliability issues in LLMs for natural language tasks, though it is incremental as it builds on existing fine-tuning approaches.
The paper tackles the problem of logical inconsistency in large language models by introducing a training objective based on probabilistic reasoning, resulting in improved consistency and better extrapolation to unseen facts compared to baselines.
Large language models (LLMs) are a promising venue for natural language understanding and generation tasks. However, current LLMs are far from reliable: they are prone to generate non-factual information and, more crucially, to contradict themselves when prompted to reason about beliefs of the world. These problems are currently addressed with large scale fine-tuning or by delegating consistent reasoning to external tools. In this work, we strive for a middle ground and introduce a training objective based on principled probabilistic reasoning that teaches a LLM to be consistent with external knowledge in the form of a set of facts and rules. Fine-tuning with our loss on a limited set of facts enables our LLMs to be more logically consistent than previous baselines and allows them to extrapolate to unseen but semantically similar factual knowledge more systematically.