UniArk: Improving Generalisation and Consistency for Factual Knowledge Extraction through Debiasing
This work addresses biases in factual knowledge extraction for language model applications, offering an incremental improvement in generalization and consistency.
The paper tackled the problem of biases in factual knowledge extraction from language models, showing that debiasing pre-training and tuning objectives improves generalization over unseen prompts. The proposed UniArk framework significantly enhanced out-of-domain generalization and consistency without extra parameters, as demonstrated through extensive experiments.
Several recent papers have investigated the potential of language models as knowledge bases as well as the existence of severe biases when extracting factual knowledge. In this work, we focus on the factual probing performance over unseen prompts from tuning, and using a probabilistic view we show the inherent misalignment between pre-training and downstream tuning objectives in language models for probing knowledge. We hypothesize that simultaneously debiasing these objectives can be the key to generalisation over unseen prompts. We propose an adapter-based framework, UniArk, for generalised and consistent factual knowledge extraction through simple methods without introducing extra parameters. Extensive experiments show that UniArk can significantly improve the model's out-of-domain generalisation as well as consistency under various prompts. Additionally, we construct ParaTrex, a large-scale and diverse dataset for measuring the inconsistency and out-of-domain generation of models. Further, ParaTrex offers a reference method for constructing paraphrased datasets using large language models.