The student becomes the master: Matching GPT3 on Scientific Factual Error Correction

Due to the prohibitively high cost of creating error correction datasets, most Factual Claim Correction methods rely on a powerful verification model to guide the correction process. This leads to a significant drop in performance in domains like Scientific Claim Correction, where good verification models do not always exist. In this work, we introduce a claim correction system that makes no domain assumptions and does not require a verifier but is able to outperform existing methods by an order of magnitude -- achieving 94% correction accuracy on the SciFact dataset, and 62.5% on the SciFact-Open dataset, compared to the next best methods 0.5% and 1.50% respectively. Our method leverages the power of prompting with LLMs during training to create a richly annotated dataset that can be used for fully supervised training and regularization. We additionally use a claim-aware decoding procedure to improve the quality of corrected claims. Our method is competitive with the very LLM that was used to generate the annotated dataset -- with GPT3.5 achieving 89.5% and 60% correction accuracy on SciFact and SciFact-Open, despite using 1250 times as many parameters as our model.