ResearchBenchmarksText-to-SQL generation on OODFollow40Overall Match AccuracyQwen 2.53.28812.81922.3531.881Jun 18, 2026Evaluation ResultsMethodMethodLinksOverall Match AccuracyQwen 2.5Params=7BParams=7B2026.0640Qwen 2.5Params=14BParams=14B2026.0640Gemma 2Params=9BParams=9B2026.0635Gemma 2Params=27BParams=27B2026.0634.3Gemma 2Params=2BParams=2B2026.0630.3Qwen 2.5Params=0.5BParams=0.5B2026.064.7