| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | DeepMind 70B Model (SFT+ORM-RL, ORM reranking) | – | Accuracy: 87.3Parameters (Billion): 70 |
| arithmetic-reasoning-on-gsm8k | DeepMind 70B Model (SFT+PRM-RL, PRM reranking) | – | Accuracy: 87.1Parameters (Billion): 70 |
| arithmetic-reasoning-on-gsm8k | DeepMind 70B Model (STaR, maj1@96) | – | Accuracy: 76.5Parameters (Billion): 70 |