Code Generation on Polyglot full-benchmark
30.7Pass RateBest DGM agent
Evaluation Results
| Method | Links | |
|---|---|---|
| Best DGM agentAgent Stage=Best, Evaluation Protocol=Endpoint2026.06 | 30.7 | |
| Best DGM agentAgent Stage=Best, Evaluation Protocol=Transfer (from SWE-bench)2026.06 | 28.9 | |
| Initial agentAgent Stage=Initial, Evaluation Protocol=Endpoint2026.06 | 14.2 | |
| Initial agentAgent Stage=Initial, Evaluation Protocol=Transfer (from SWE-bench)2026.06 | 14.2 |