ResearchBenchmarksOperational Metrics for LLM Fine-tuning on SST-2 (test)Follow16,752Number of QueriesMeZO702.724,869.369,03613,202.64Nov 11, 2025Evaluation ResultsMethodMethodLinksNumber of QueriesFine-tuning TimeMeZOBackbone=GPT-2-XLBackbone=GPT-2-XL2025.1116,7523.13LOZOBackbone=GPT-2-XLBackbone=GPT-2-XL2025.1110,6860.66MeZO-SVRGBackbone=GPT-2-XLBackbone=GPT-2-XL2025.114,2481.51HiZOOBackbone=GPT-2-XLBackbone=GPT-2-XL2025.112,2320.98MeZO-AdamBackbone=GPT-2-XLBackbone=GPT-2-XL2025.111,6320.32LORENBackbone=GPT-2-XLBackbone=GPT-2-XL2025.111,3200.33