Social Interaction Simulation on EduMirror social interaction steps (144 steps) (test)
4.958NaturalnessEduMirror
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EduMirrorLLM Backbone=GPT-4.12026.06 | 4.958 | 4.958 | |
| EduMirrorLLM Backbone=Average2026.06 | 4.802 | 4.821 | |
| EduMirrorLLM Backbone=Qwen32026.06 | 4.792 | 4.824 | |
| EduMirrorLLM Backbone=DeepSeek2026.06 | 4.75 | 4.792 | |
| EduMirrorLLM Backbone=Gemini2026.06 | 4.708 | 4.708 | |
| ReActLLM Backbone=GPT-4.12026.06 | 4.667 | 4.86 | |
| LLMobLLM Backbone=GPT-4.12026.06 | 4.625 | 4.875 | |
| LLMobLLM Backbone=Average2026.06 | 4.229 | 4.469 | |
| ReActLLM Backbone=Gemini2026.06 | 4.208 | 4.417 | |
| ReActLLM Backbone=Average2026.06 | 4.208 | 4.496 | |
| LLMobLLM Backbone=Gemini2026.06 | 4.167 | 4.292 | |
| LLMobLLM Backbone=DeepSeek2026.06 | 4.083 | 4.375 | |
| JAG-ConcordiaLLM Backbone=Qwen32026.06 | 4.083 | 4.192 | |
| D2ALLM Backbone=Gemini2026.06 | 4.042 | 4.181 | |
| LLMobLLM Backbone=Qwen32026.06 | 4.042 | 4.333 | |
| ReActLLM Backbone=DeepSeek2026.06 | 4 | 4.5 | |
| JAG-ConcordiaLLM Backbone=GPT-4.12026.06 | 3.958 | 4.133 | |
| ReActLLM Backbone=Qwen32026.06 | 3.958 | 4.208 | |
| BabyAGILLM Backbone=Qwen32026.06 | 3.958 | 3.958 | |
| D2ALLM Backbone=DeepSeek2026.06 | 3.925 | 4.1 | |
| JAG-ConcordiaLLM Backbone=Average2026.06 | 3.922 | 4.063 | |
| JAG-ConcordiaLLM Backbone=Gemini2026.06 | 3.885 | 4.052 | |
| D2ALLM Backbone=Average2026.06 | 3.884 | 4.083 | |
| BabyAGILLM Backbone=DeepSeek2026.06 | 3.875 | 4.042 | |
| D2ALLM Backbone=Qwen32026.06 | 3.867 | 4.117 | |
| JAG-ConcordiaLLM Backbone=DeepSeek2026.06 | 3.76 | 3.875 | |
| D2ALLM Backbone=GPT-4.12026.06 | 3.7 | 3.933 | |
| BabyAGILLM Backbone=Average2026.06 | 3.698 | 3.875 | |
| BabyAGILLM Backbone=Gemini2026.06 | 3.5 | 3.708 | |
| BabyAGILLM Backbone=GPT-4.12026.06 | 3.458 | 3.792 |