Binary decision task on CGSS China (test)
70.17AccuracyOG-MAR
Evaluation results
| Method | Links | |
|---|---|---|
| OG-MARBackbone=Gemini 2.5 Flash Lite2026.01 | 70.17 | |
| RoleBackbone=Gemini 2.5 Flash Lite2026.01 | 66.69 | |
| OG-MARBackbone=GPT-4o mini2026.01 | 65.09 | |
| Zero-shotBackbone=Gemini 2.5 Flash Lite2026.01 | 64.67 | |
| DebateBackbone=Gemini 2.5 Flash Lite2026.01 | 63.48 | |
| OG-MARBackbone=EXAONE 3.52026.01 | 63.07 | |
| ValuesRAGBackbone=Gemini 2.5 Flash Lite2026.01 | 60.84 | |
| Self-consistencyBackbone=Gemini 2.5 Flash Lite2026.01 | 60.63 | |
| RoleBackbone=GPT-4o mini2026.01 | 60.14 | |
| DebateBackbone=GPT-4o mini2026.01 | 59.93 | |
| ValuesRAGBackbone=GPT-4o mini2026.01 | 58.89 | |
| Zero-shotBackbone=GPT-4o mini2026.01 | 58.47 | |
| ValuesRAGBackbone=EXAONE 3.52026.01 | 58.33 | |
| Self-consistencyBackbone=GPT-4o mini2026.01 | 56.31 | |
| DebateBackbone=EXAONE 3.52026.01 | 56.24 | |
| OG-MARBackbone=QWEN 2.52026.01 | 52.2 | |
| ValuesRAGBackbone=QWEN 2.52026.01 | 46.97 | |
| DebateBackbone=QWEN 2.52026.01 | 45.78 | |
| RoleBackbone=QWEN 2.52026.01 | 34.63 | |
| Self-consistencyBackbone=QWEN 2.52026.01 | 32.89 | |
| RoleBackbone=EXAONE 3.52026.01 | 31.29 | |
| Zero-shotBackbone=EXAONE 3.52026.01 | 28.85 | |
| Zero-shotBackbone=QWEN 2.52026.01 | 27.04 | |
| Self-consistencyBackbone=EXAONE 3.52026.01 | 26.97 |