In-Context Value Alignment on Value Composition (Overall)
3.995Confucianism ScoreQ+IF+COT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Q+IF+COTBackbone Model=Gemini-1.5-Flash, Evaluator=DeepSeek-V3.12025.07 | 3.995 | 3.339 | 4.27 | 4.461 | 3.891 | |
| OPROBackbone Model=Gemini-1.5-Flash, Evaluator=DeepSeek-V3.12025.07 | 3.885 | 3.098 | 4.348 | 4.377 | 4.187 | |
| PICACOTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.788 | 3.135 | 4.257 | 4.287 | 4.173 | |
| OPROTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.713 | 2.961 | 4.286 | 4.287 | 4.153 | |
| PICACOBackbone Model=GPT-3.5-Turbo, Evaluator=DeepSeek-V3.12025.07 | 3.7 | 3.061 | 4.198 | 4.248 | 4.086 | |
| Q+IF+COTBackbone Model=GPT-3.5-Turbo, Evaluator=DeepSeek-V3.12025.07 | 3.691 | 2.933 | 4.082 | 4.184 | 3.835 | |
| PICACOBackbone Model=Gemini-1.5-Flash, Evaluator=DeepSeek-V3.12025.07 | 3.666 | 2.876 | 4.359 | 4.397 | 4.273 | |
| OPROBackbone Model=GPT-3.5-Turbo, Evaluator=DeepSeek-V3.12025.07 | 3.66 | 2.892 | 4.249 | 4.288 | 4.036 | |
| MODULAR PLURALISMBackbone Model=Gemini-1.5-Flash, Evaluator=DeepSeek-V3.12025.07 | 3.648 | 2.805 | 4.017 | 4.423 | 3.936 | |
| MODULAR PLURALISMBackbone Model=GPT-3.5-Turbo, Evaluator=DeepSeek-V3.12025.07 | 3.625 | 2.672 | 4.205 | 4.17 | 4.104 | |
| URIALTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.622 | 3.03 | 4.097 | 4.164 | 4.087 | |
| URIAL+SUMBackbone Model=Gemini-1.5-Flash, Evaluator=DeepSeek-V3.12025.07 | 3.569 | 3 | 4.135 | 4.262 | 4.008 | |
| MODULARTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.567 | 3.036 | 4.245 | 4.236 | 4.177 | |
| URIAL+SUMTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.559 | 2.942 | 4.046 | 4.106 | 4.046 | |
| URIALTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.53 | 3.03 | 4.085 | 4.079 | 4.073 | |
| URIAL+SUMBackbone Model=GPT-3.5-Turbo, Evaluator=DeepSeek-V3.12025.07 | 3.507 | 2.73 | 3.977 | 3.985 | 3.848 | |
| PICACOTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.471 | 2.987 | 4.11 | 4.118 | 4.112 | |
| OPROTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.437 | 2.925 | 4.114 | 4.104 | 4.095 | |
| MODULARTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.427 | 2.973 | 3.793 | 4.025 | 3.94 | |
| PICACOBackbone Model=LLaMA-3.1-8B-Instruct, Evaluator=DeepSeek-V3.12025.07 | 3.424 | 2.586 | 4.24 | 4.187 | 4.148 | |
| QTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.411 | 2.661 | 3.891 | 3.891 | 3.891 | |
| CICLTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.372 | 2.842 | 4.077 | — | — | |
| Q+IF+COTBackbone Model=LLaMA-3.1-8B-Instruct, Evaluator=DeepSeek-V3.12025.07 | 3.349 | 2.762 | 4.092 | 4.077 | 4.033 | |
| MP+SYSTEM 2Target LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.331 | 2.482 | 3.914 | 3.746 | 3.888 | |
| MP+SYSTEM 2Target LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.324 | 2.468 | 4.227 | 4.245 | 4.124 | |
| URIAL+SUMTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.316 | 2.898 | 3.921 | 3.922 | 3.921 | |
| Q+IFTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.306 | 2.728 | 4.082 | 4.247 | 4.032 | |
| QTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.288 | 2.671 | 3.978 | 3.978 | 3.978 | |
| CICLTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.238 | 2.703 | 3.909 | — | — | |
| Q+IFTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.164 | 2.653 | 3.977 | 3.952 | 3.989 | |
| MODULAR PLURALISMBackbone Model=LLaMA-3.1-8B-Instruct, Evaluator=DeepSeek-V3.12025.07 | 3.146 | 2.492 | 3.952 | 4.118 | 4.055 | |
| Q+IF+CoTTarget LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.132 | 2.646 | 3.986 | 4.01 | 3.951 | |
| MP+SYSTEM 1Target LLM=LLaMA-3.1-8B-Instruct, Judge Model=GPT-4o2025.07 | 3.105 | 2.401 | 3.972 | 3.896 | 3.955 | |
| Q+IF+CoTTarget LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 3.034 | 2.719 | 4.068 | 4.189 | 4.01 | |
| OPROBackbone Model=LLaMA-3.1-8B-Instruct, Evaluator=DeepSeek-V3.12025.07 | 3.013 | 2.571 | 4.181 | 4.201 | 4.164 | |
| MP+SYSTEM 1Target LLM=GPT-3.5-Turbo, Judge Model=GPT-4o2025.07 | 2.983 | 2.058 | 4.058 | 4.14 | 3.927 | |
| URIAL+SUMBackbone Model=LLaMA-3.1-8B-Instruct, Evaluator=DeepSeek-V3.12025.07 | 2.957 | 2.387 | 3.848 | 3.799 | 3.784 |