Domain-level accuracy estimation on STEM benchmarks (GSM8K, OlympiadBench, MATH, SciBench, etc.) (held-out)
100External ScorePHI-3.5-MINI
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| PHI-3.5-MINIParameters=3B2026.01 | 100 | 95 | 95 | 94 | 90 | 90 | 88 | 90 | 89 | 90 | |
| MISTRALParameters=7B2026.01 | 96 | 79 | 96 | 95 | 92 | 68 | 61 | 66 | 55 | 68 | |
| MIXTRALParameters=8x7B2026.01 | 96 | 89 | 98 | 96 | 96 | 66 | 47 | 62 | 73 | 73 | |
| QWEN2Parameters=7B2026.01 | 95 | 94 | 96 | 98 | 95 | 95 | 96 | 95 | 98 | 95 | |
| LLAMA-3Parameters=8B2026.01 | 95 | 92 | 95 | 95 | 55 | 65 | 90 | 65 | 100 | 65 | |
| GEMMAParameters=2B2026.01 | 94 | 92 | 89 | 89 | 87 | 87 | 89 | 87 | 98 | 87 | |
| GEMMAParameters=7B2026.01 | 92 | 85 | 90 | 88 | 81 | 84 | 93 | 84 | 95 | 83 | |
| GPT-NEOXParameters=20B2026.01 | 90 | 89 | 94 | 98 | 82 | 58 | 37 | 50 | 70 | 50 | |
| QWEN2Parameters=72B2026.01 | 76 | 75 | 93 | 79 | 75 | 75 | 89 | 75 | 95 | 75 |