ResearchBenchmarksGrammar Synthesis on L2 a^n b^n c^mFollow100AccuracyLlama 1B2.96828.15953.3578.541Apr 12, 2026Evaluation ResultsMethodMethodLinksAccuracyLlama 1BG=GASGG=GASG2026.04100Llama 1BG=G^ ASGG=G^ ASG2026.04100Llama 70BG=GASGG=GASG2026.04100Llama 70BG=G^ ASGG=G^ ASG2026.04100o1G=-G=-2026.0496.7o4 miniG=-G=-2026.0493.3o3 miniG=-G=-2026.0486.7DeepSeek-R1G=-G=-2026.0486.7GPT 4.1G=-G=-2026.0476.7Llama 70BG=-G=-2026.0453.3Llama 1BG=-G=-2026.046.7