Code Generation Benchmarks Textual
84.1HumanEval+Uni-OPD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Uni-OPDStudent Model=Qwen3-VL-4B-Instruct2026.05 | 84.1 | 71.4 | 41.3 | 65.6 | |
| OPDStudent Model=Qwen3-VL-4B-Instruct2026.05 | 83.1 | 70.6 | 38.6 | 64.1 | |
| TeacherOrigin=Domain-specific RL on textual code, Teacher Model=Qwen3-VL-4B-Instruct-Code-RL2026.05 | 82.2 | 70.5 | 40.1 | 64.3 | |
| StudentStudent Model=Qwen3-VL-4B-Instruct2026.05 | 76.8 | 70 | 37 | 61.3 |