ResearchBenchmarksPremise Generation on Crafted Rules Dataset (test)Follow0.411BLEUEngine0.070920.159210.24750.33579Feb 18, 2024Evaluation ResultsMethodMethodLinksBLEUSelf-BLEUFact CountEnginebackbone=Mistral-7b, m...backbone=Mistral-7b, mode=fine-tuned2024.020.4110.6873.42GPT-4mode=promptingmode=prompting2024.020.1490.8052.58GPT-3.5mode=promptingmode=prompting2024.020.0840.7391.72