ResearchBenchmarksPremise Completion on Crafted Rules Dataset (test)Follow52.7BLEUEngine16.50825.90435.344.696Feb 18, 2024Evaluation ResultsMethodMethodLinksBLEUEnginebackbone=Mistral-7b, m...backbone=Mistral-7b, mode=fine-tuned2024.0252.7GPT-3.5mode=promptingmode=prompting2024.0224.8GPT-4mode=promptingmode=prompting2024.0217.9