Llemma: An Open Language Model For Mathematics

Benchmark Model Rank Results
automated-theorem-proving-on-minif2f-testLLEMMA-7b#13cumulative: 26.2Pass@32: 26.2ITP: Lean
automated-theorem-proving-on-minif2f-testLLEMMA-34b#14cumulative: 25.8Pass@32: 25.8ITP: Lean
arithmetic-reasoning-on-gsm8kLlemma 34B#234Accuracy: 51.5Parameters (Billion): 34
arithmetic-reasoning-on-gsm8kLlemma 7B#251Accuracy: 36.4Parameters (Billion): 7