Calibration
SciCode → Vals LiveCodeBench
Vals LiveCodeBench is estimated from SciCode with a linear curve fitted on 7 models measured on both: y = 0.2490·x + 0.7387, R² = 0.75, cross-validated error 0.8 pp. It is used for 0 estimates.
Fitted on
| Model | SciCode | Vals LiveCodeBench |
|---|---|---|
| Gemini 3.5 Flash | 53.1% | 87.6% |
| Grok 4.3 | 47.3% | 84.5% |
| Inkling-Small | 48.7% | 85.9% |
| Kimi K2.6 | 52.2% | 86.8% |
| Ling 3.0 Flash | 41.2% | 84.0% |
| Nemotron 3 Ultra | 44.6% | 86.0% |
| Qwen3.7 Max | 53.5% | 87.1% |