End-to-end pipeline generation on 150 ML tasks dataset-goal pairs (test)
98Success RateManual Construction
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Manual ConstructionConfiguration=Manual Construction2026.04 | 98 | 3,240 | 5,890 | 84 | 0.21 | |
| Full SystemConfiguration=Full System2026.04 | 84.7 | 143 | 487 | 81 | 0.24 | |
| Ablation-NoHistoryConfiguration=Ablation-NoHistory2026.04 | 79.3 | 156 | 521 | 79 | 0.26 | |
| Baseline-SemanticOnlyConfiguration=Baseline-SemanticOnly2026.04 | 71.3 | 167 | 589 | 77 | 0.27 | |
| Baseline-DocBasedConfiguration=Baseline-DocBased2026.04 | 68 | 189 | 673 | 76 | 0.29 |