Text-based comment generation on Hot-Comment (test)
21.08BLEU-1StyleCmt
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| StyleCmtBase Model=Qwen2.5-14B2026.04 | 21.08 | 18.61 | 63.18 | 60.78 | 82.18 | 84.5 | |
| StyleCmtBase Model=ChatGPT-4o2026.04 | 20.57 | 21.15 | 63.98 | 59.49 | 76.98 | 77.69 | |
| StyleCmtBase Model=LLaMA3.1-8B2026.04 | 20 | 20.48 | 60.9 | 57.74 | 73.52 | 71.49 | |
| Qwen2.5-7B2026.04 | 17.08 | 17.14 | 58.63 | 45.17 | 75.84 | 76.71 | |
| LLaMA3.1-8B2026.04 | 17.03 | 16.64 | 57.78 | 47.66 | 47.25 | 63.99 | |
| ChatGPT-4o2026.04 | 16.97 | 16.41 | 59.78 | 57.09 | 71.41 | 72.38 | |
| Qwen2.5-14B2026.04 | 15.36 | 15.64 | 58.02 | 51.78 | 76.7 | 70.19 | |
| Qwen2.5-0.5B2026.04 | 14.22 | 13.41 | 57.79 | 31.34 | 66.69 | 62.23 | |
| Mistral-7B2026.04 | 8.11 | 9.75 | 53.89 | 49.58 | 69.63 | 63.71 | |
| Baichuan2-7B2026.04 | 7.34 | 8.94 | 57.04 | 39.58 | 57.39 | 31.68 |