Programming Language Generation on Q programming language (test)
89.1Valid QGRPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GRPOBackbone=QQwen-7B-Pretrain, Samples=5, Evaluation Protocol=standard GRPO training2026.03 | 89.1 | 27.3 | 5 | |
| CBRL GRPOBackbone=QQwen-7B-Pretrain, Samples=5, Evaluation Protocol=context bootstrapping augmented GRPO2026.03 | 80.9 | 43 | 26.3 | |
| Baseline w/ Few ShotsBackbone=QQwen-7B-Pretrain, Samples=5, Evaluation Protocol=few-shot prompting2026.03 | 4.3 | 0.6 | 0.4 | |
| BaselineBackbone=QQwen-7B-Pretrain, Samples=5, Evaluation Protocol=zero-shot2026.03 | 3.1 | 0.3 | 0 |