Your Finetuned Large Language Model is Already a Powerful Out-of-distribution Detector

TL;DR

This paper proposes using the likelihood ratio between pretrained and finetuned large language models as an effective, easy-to-implement criterion for out-of-distribution detection across various tasks, enhancing model reliability.

Contribution

It introduces a novel OOD detection method based on likelihood ratios of LLMs that requires no additional training and can be applied directly using existing models and frameworks.

Findings

Likelihood ratio effectively detects OOD data in multiple scenarios.

The method is simple to implement with existing neural network loss functions.

Evaluation shows strong performance across diverse OOD detection tasks.

Abstract

We revisit the likelihood ratio between a pretrained large language model (LLM) and its finetuned variant as a criterion for out-of-distribution (OOD) detection. The intuition behind such a criterion is that, the pretrained LLM has the prior knowledge about OOD data due to its large amount of training data, and once finetuned with the in-distribution data, the LLM has sufficient knowledge to distinguish their difference. Leveraging the power of LLMs, we show that, the likelihood ratio can serve as an effective OOD detection criterion. Moreover, we apply the proposed LLM-based likelihood ratio to detect OOD questions in question-answering (QA) systems, which can be used to improve the performance of specialized LLMs for general questions. Given that likelihood can be easily obtained by the loss functions within contemporary neural network frameworks, it is straightforward to implement this approach in practice. Since both the pretrained LLMs and its various finetuned models are widely available from online platforms such as Hugging Face, our proposed criterion can be effortlessly incorporated for OOD detection without the need for further training. We conduct comprehensive evaluation across on multiple settings, including far OOD, near OOD, spam detection, and QA scenarios, to demonstrate the effectiveness of the method. Code can be found at https://github.com/andiac/LLMOODratio