| Arm | Shipped & correct, no feedback | Shipped & correct, with ≤2 compiler-feedback rounds | 95% CI |
|---|---:|---:|---|
| python | 89/152 | 90/152 (59.2%) | 51.3%–66.7% |
| python+mypy | 86/152 | 94/152 (61.8%) | 53.9%–69.2% |
| javascript | 86/152 | 87/152 (57.2%) | 49.3%–64.8% |
| typescript | 53/152 | 93/152 (61.2%) | 53.2%–68.6% |
| java | 75/152 | 86/152 (56.6%) | 48.6%–64.2% |
| go | 49/152 | 75/152 (49.3%) | 41.5%–57.2% |
| rust | 57/152 | 60/152 (39.5%) | 32.0%–47.4% |

| Arm | First-try pass | Rejected by gate (first try) | Wrong first tries caught by gate | Final pass after ≤2 compiler-feedback rounds | Gained / lost | Compiles but wrong (final) | Pass rate when gate is green |
|---|---:|---:|---:|---:|---:|---:|---:|
| python | 89/152 (58.6%) | 3 | 3/63 (5%) | 90/152 (59.2%) | +1 / −0 | 61 | 59.6% |
| python+mypy | 89/152 (58.6%) | 13 | 10/63 (16%) | 95/152 (62.5%) | +6 / −0 | 54 | 64.2% |
| javascript | 86/152 (56.6%) | 3 | 3/66 (5%) | 87/152 (57.2%) | +1 / −0 | 64 | 57.6% |
| typescript | 56/152 (36.8%) | 63 | 60/96 (62%) | 96/152 (63.2%) | +40 / −0 | 52 | 66.2% |
| java | 75/152 (49.3%) | 36 | 36/77 (47%) | 86/152 (56.6%) | +11 / −0 | 52 | 62.3% |
| go | 49/152 (32.2%) | 68 | 68/103 (66%) | 75/152 (49.3%) | +26 / −0 | 54 | 58.1% |
| rust | 57/152 (37.5%) | 53 | 53/95 (56%) | 60/152 (39.5%) | +3 / −0 | 49 | 55.0% |