Code Generation on LiveCodeBench v6 (Pass rate)
61.5Pass RateEVOTD
Evaluation Results
| Method | Links | |
|---|---|---|
| EVOTDEvaluation Protocol=Pass@8, Training Iteration=Iter3, Backbone=Qwen3-4B (Thinking)2026.05 | 61.5 | |
| EVOTDEvaluation Protocol=Pass@8, Training Iteration=Iter2, Backbone=Qwen3-4B (Thinking)2026.05 | 60.8 | |
| EVOTDEvaluation Protocol=Pass@8, Training Iteration=Iter1, Backbone=Qwen3-4B (Thinking)2026.05 | 58.9 | |
| Evol-InstructEvaluation Protocol=Pass@8, Backbone=Qwen3-4B (Thinking)2026.05 | 58 | |
| Qwen3-4B (Thinking)Evaluation Protocol=Pass@82026.05 | 53.8 | |
| EVOTDEvaluation Protocol=Pass@1, Training Iteration=Iter3, Backbone=Qwen3-4B (Thinking)2026.05 | 49.8 | |
| EVOTDEvaluation Protocol=Pass@1, Training Iteration=Iter2, Backbone=Qwen3-4B (Thinking)2026.05 | 48.5 | |
| Evol-InstructEvaluation Protocol=Pass@1, Backbone=Qwen3-4B (Thinking)2026.05 | 47.4 | |
| EVOTDEvaluation Protocol=Pass@1, Training Iteration=Iter1, Backbone=Qwen3-4B (Thinking)2026.05 | 46.8 | |
| Qwen3-4B (Thinking)Evaluation Protocol=Pass@12026.05 | 42.9 |