GSM8K (Maj1@32)
Math · 2023-12-19
GSM8K accuracy with chain-of-thought self-consistency: the majority answer over 32 samples. Kept separate from single-sample GSM8K.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 1.0 Ultra | 94.4 |
| Gemini 1.0 Pro | 86.5 |
Benchmarks
218 sources · 1,060 models × 2,130 benchmarks · 109,661 scoresMath · 2023-12-19
GSM8K accuracy with chain-of-thought self-consistency: the majority answer over 32 samples. Kept separate from single-sample GSM8K.
| Model | Score |
|---|---|
| Gemini 1.0 Ultra | 94.4 |
| Gemini 1.0 Pro | 86.5 |