Moral Alignment on Moral Alignment
89.8Justice F1GPT-4omini
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-4ominiConsensus Setting=Five LLMs2025.06 | 89.8 | 83.91 | 80.18 | 78.04 | 82.31 | |
| GPT-4ominiConsensus Setting=Four LLMs2025.06 | 88.73 | 83.01 | 78.57 | 78.02 | 81.81 | |
| MoonshotConsensus Setting=Five LLMs2025.06 | 83.47 | 80.31 | 58.9 | 78.84 | 72.75 | |
| ClaudeConsensus Setting=Four LLMs2025.06 | 75.78 | 67.56 | 74.52 | 78.2 | 60.4 | |
| Llama2-13BConsensus Setting=Four LLMs, Fine-tuned token embeddings=Deontology and Utilitarianism2025.06 | 75.25 | 63.37 | 37.68 | 41.55 | 45.06 | |
| Llama2-13BConsensus Setting=Five LLMs, Fine-tuned token embeddings=Deontology and Utilitarianism2025.06 | 75 | 63.2 | 38.84 | 40.84 | 45.47 | |
| ClaudeConsensus Setting=Five LLMs2025.06 | 74.27 | 65.84 | 73.31 | 76.11 | 58.12 | |
| GPT-3.5Consensus Setting=Four LLMs2025.06 | 74.05 | 77.13 | 56.49 | 65.86 | 68.29 | |
| GPT-3.5Consensus Setting=Five LLMs2025.06 | 72.46 | 75.52 | 57.7 | 64.26 | 66.08 |