Calibration
ARC-AGI-2 → LABBench2
LABBench2 is estimated from ARC-AGI-2 with a offset logistic curve fitted on 5 models measured on both: y = 0.8070 + (0.8417 − 0.8070) / (1 + exp(−200.00·(x − 0.8480))), R² = 0.49, cross-validated error 2.5 pp. It is used for 3 estimates.
| Estimated model | ARC-AGI-2 | LABBench2 | Source |
|---|---|---|---|
| Claude Sonnet 4.5 | 13.6% | 80.7% | estimated ± 2.5 pp, low confidence |
| dots3-note Preview | 81.4% | 80.7% | estimated ± 2.5 pp, low confidence |
| Gemini 3 Pro Deep Think | 45.1% | 80.7% | estimated ± 2.5 pp, low confidence |