Writing capabilities on WritingBench (test)
8.56ScoreTCR
Evaluation Results
| Method | Links | |
|---|---|---|
| TCROptimizer=GRPO, Foundation Model=Qwen3-14B2025.10 | 8.56 | |
| PREFOptimizer=GRPO, Foundation Model=Qwen3-14B2025.10 | 8.44 | |
| PREFOptimizer=GSPO, Foundation Model=Qwen3-8B2025.10 | 8.39 | |
| TCROptimizer=GSPO, Foundation Model=Qwen3-8B2025.10 | 8.37 | |
| ELOOptimizer=GRPO, Foundation Model=Qwen3-14B2025.10 | 8.32 | |
| LWOptimizer=GRPO, Foundation Model=Qwen3-14B2025.10 | 8.24 | |
| ELOOptimizer=GSPO, Foundation Model=Qwen3-8B2025.10 | 8.19 | |
| PWOptimizer=GRPO, Foundation Model=Qwen3-14B2025.10 | 8.12 | |
| LWOptimizer=GSPO, Foundation Model=Qwen3-8B2025.10 | 8.11 | |
| Base ModelOptimizer=GRPO, Foundation Model=Qwen3-14B2025.10 | 7.95 | |
| PWOptimizer=GSPO, Foundation Model=Qwen3-8B2025.10 | 7.86 | |
| Base ModelOptimizer=GSPO, Foundation Model=Qwen3-8B2025.10 | 7.65 |