Violated Principle Classification on ETHICSCAFF
94.55AccuracyQwen3-14B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-14BModel Family=Qwen3-14B, Variant=w/ ETHICSCAFF (Full)2025.08 | 94.55 | 95.21 | 94.55 | 94.37 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=w/ ETHICSCAFF (Full)2025.08 | 88.98 | 91.11 | 88.98 | 88.63 | |
| Qwen3-14BModel Family=Qwen3-14B, Variant=w/ ETHICSCAFF (SFT)2025.08 | 74.33 | 75.27 | 74.33 | 70.39 | |
| GPT-4oModel Family=Closed-Source Models, Variant=RAG2025.08 | 73.61 | 81.52 | 73.61 | 73.12 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=w/ ETHICSCAFF (SFT)2025.08 | 73.61 | 82.17 | 73.61 | 75.2 | |
| Qwen3-14BModel Family=Qwen3-14B, Variant=RAG2025.08 | 68.48 | 77.69 | 68.48 | 68.4 | |
| Llama3.1-8BModel Family=Llama3.1-8B, Variant=w/ ETHICSCAFF2025.08 | 67.96 | 63.22 | 67.96 | 62.22 | |
| Qwen3-14BModel Family=Qwen3-14B, Variant=Few-shot + CoT2025.08 | 64.15 | 70.99 | 64.15 | 59.87 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=Few-shot + CoT2025.08 | 57.62 | 58.95 | 57.62 | 51.55 | |
| Llama3.1-8BModel Family=Llama3.1-8B, Variant=RAG2025.08 | 57.22 | 62.27 | 57.22 | 54.98 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=RAG2025.08 | 55.02 | 63.51 | 55.02 | 52.02 | |
| Claude-Sonnet-4Model Family=Closed-Source Models, Variant=Base2025.08 | 54.59 | 65.82 | 54.59 | 47.44 | |
| Qwen3-14BModel Family=Qwen3-14B, Variant=Few-shot2025.08 | 52.24 | 61.59 | 52.24 | 47.04 | |
| DeepSeek-V3-685BModel Family=Open-Source Models, Variant=Base2025.08 | 52.14 | 64.04 | 52.14 | 46.85 | |
| GPT-4oModel Family=Closed-Source Models, Variant=Base2025.08 | 51.18 | 62.69 | 51.18 | 45.66 | |
| Qwen3-14BModel Family=Qwen3-14B, Variant=CoT2025.08 | 50.94 | 59.34 | 50.94 | 46.92 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=CoT2025.08 | 50.22 | 57.03 | 50.22 | 44.8 | |
| Qwen3-14BModel Family=Qwen3-14B, Variant=Base2025.08 | 42.31 | 66.35 | 42.31 | 38.67 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=Few-shot2025.08 | 33.55 | 46.13 | 33.55 | 31.58 | |
| Llama3.1-8BModel Family=Llama3.1-8B, Variant=Base2025.08 | 29.39 | 43.73 | 29.39 | 25.96 | |
| Qwen3-8BModel Family=Qwen3-8B, Variant=Base2025.08 | 23.33 | 44.16 | 23.33 | 22.18 | |
| Llama3.1-8BModel Family=Llama3.1-8B, Variant=Few-shot2025.08 | 18.12 | 22 | 18.12 | 16.07 |