Diversity and Quality Evaluation on NoveltyBench
7.13LLM DiversityDiverse Outline
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Diverse OutlineLv.=2, Backbone=GPT-5-mini2026.06 | 7.13 | 0.5 | 0.35 | 16.75 | 7.39 | |
| KeywordLv.=2, Backbone=GPT-5-mini2026.06 | 6.63 | 0.43 | 0.31 | 16.46 | 7.32 | |
| KeywordLv.=2, Backbone=Qwen3-4B2026.06 | 5.01 | 0.38 | 0.25 | 13.21 | 7.4 | |
| Diverse OutlineLv.=2, Backbone=Qwen3-4B2026.06 | 5 | 0.41 | 0.27 | 13.85 | 7.09 | |
| VerbalizedLv.=0, Backbone=GPT-5-mini2026.06 | 3.1 | 0.36 | 0.22 | 9.76 | 7.3 | |
| DirectLv.=0, Backbone=GPT-5-mini2026.06 | 2.88 | 0.41 | 0.24 | 15.25 | 7.3 | |
| VerbalizedLv.=0, Backbone=Qwen3-4B2026.06 | 2.77 | 0.37 | 0.18 | 11.01 | 6.26 | |
| G2Lv.=0, Backbone=Qwen3-4B2026.06 | 2.75 | 0.21 | 0.23 | 5.54 | 7.35 | |
| DirectLv.=0, Backbone=Qwen3-4B2026.06 | 2.44 | 0.3 | 0.18 | 10.96 | 6.72 | |
| SSOTLv.=1, Backbone=GPT-5-mini2026.06 | 2.36 | 0.24 | 0.19 | 4.74 | 6.88 | |
| DirectLv.=0, Backbone=Qwen3-4B, Temperature (τ)=1.52026.06 | 2.36 | 0.25 | 0.17 | 7.38 | 6.51 | |
| SSOTLv.=1, Backbone=Qwen3-4B2026.06 | 2.13 | 0.2 | 0.17 | 6.03 | 6.36 |