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

arXiv:2404.08679v1 Announce Type: cross
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, for the …

abstract arxiv criterion cs.ai cs.cl cs.lg data detection distribution intuition knowledge language language model large language large language model likelihood llm prior stat.ml type