Chat Generation on Infinite Chats
7.65LLM DiversityDiv Outline
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Div OutlineBackbone LLM=Mistral-L3, Injection Level (Lv)=22026.06 | 7.65 | 40 | 23 | 7.86 | |
| Diverse OutlineLv.=2, Backbone=GPT-5-mini2026.06 | 7.32 | 40 | 31 | 8.28 | |
| Div OutlineBackbone LLM=DeepSeek-V3.2, Injection Level (Lv)=22026.06 | 7.07 | 44 | 27 | 7.36 | |
| KeywordBackbone LLM=Mistral-L3, Injection Level (Lv)=22026.06 | 6.51 | 35 | 19 | 8.02 | |
| KeywordBackbone LLM=DeepSeek-V3.2, Injection Level (Lv)=22026.06 | 5.88 | 40 | 25 | 7.5 | |
| KeywordLv.=2, Backbone=Qwen3-4B2026.06 | 5.77 | 33 | 22 | 7.71 | |
| Diverse OutlineLv.=2, Backbone=Qwen3-4B2026.06 | 5.58 | 35 | 24 | 7.45 | |
| KeywordLv.=2, Backbone=GPT-5-mini2026.06 | 4.87 | 32 | 25 | 8.36 | |
| DirectBackbone LLM=DeepSeek-V3.2, Injection Level (Lv)=02026.06 | 3.37 | 29 | 21 | 7.51 | |
| DirectBackbone LLM=DeepSeek-V3.2, Injection Level (Lv)=0, Temperature (τ)=1.52026.06 | 3.36 | 32 | 21 | 7.07 | |
| VerbalizedBackbone LLM=DeepSeek-V3.2, Injection Level (Lv)=02026.06 | 3.03 | 31 | 19 | 6.88 | |
| SSOTBackbone LLM=DeepSeek-V3.2, Injection Level (Lv)=12026.06 | 2.86 | 26 | 19 | 6.97 | |
| VerbalizedLv.=0, Backbone=GPT-5-mini2026.06 | 2.84 | 23 | 20 | 8.47 | |
| SSOTBackbone LLM=Mistral-L3, Injection Level (Lv)=12026.06 | 2.78 | 21 | 12 | 7.26 | |
| DirectLv.=0, Backbone=GPT-5-mini2026.06 | 2.69 | 26 | 21 | 8.3 | |
| VerbalizedBackbone LLM=Mistral-L3, Injection Level (Lv)=02026.06 | 2.63 | 26 | 12 | 6.91 | |
| SSOTLv.=1, Backbone=GPT-5-mini2026.06 | 2.4 | 18 | 17 | 8.3 | |
| DirectLv.=0, Backbone=Qwen3-4B, Temperature (τ)=1.52026.06 | 2.35 | 19 | 13 | 6.67 | |
| DirectBackbone LLM=Mistral-L3, Injection Level (Lv)=02026.06 | 2.33 | 28 | 13 | 6.92 | |
| VerbalizedLv.=0, Backbone=Qwen3-4B2026.06 | 2.31 | 25 | 13 | 6.55 | |
| SSOTLv.=1, Backbone=Qwen3-4B2026.06 | 2.21 | 15 | 15 | 6.7 | |
| DirectLv.=0, Backbone=Qwen3-4B2026.06 | 2.11 | 19 | 13 | 7.02 | |
| G2Lv.=0, Backbone=Qwen3-4B2026.06 | 2.09 | 12 | 19 | 7.03 |