Discrimination on PsyCLIENT-CP Human Client
100Accuracy (A)Qwen2.5-72B-Instruct
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-72B-InstructLLM-as-a-judge=true, Binary choice evaluation=true, Positional bias mitigation=Label order reversal2026.01 | 100 | 100 | |
| GPT-4oLLM-as-a-judge=true, Binary choice evaluation=true, Positional bias mitigation=Label order reversal2026.01 | 98.1 | 98.6 | |
| DeepSeek-V3-0324LLM-as-a-judge=true, Binary choice evaluation=true, Positional bias mitigation=Label order reversal2026.01 | 97.8 | 100 | |
| Claude-Sonet-3.5LLM-as-a-judge=true, Binary choice evaluation=true, Positional bias mitigation=Label order reversal2026.01 | 95 | 98.6 | |
| Qwen3-235B-A22BLLM-as-a-judge=true, Binary choice evaluation=true, Positional bias mitigation=Label order reversal2026.01 | 79.4 | 78.6 | |
| DeepSeek-R1LLM-as-a-judge=true, Binary choice evaluation=true, Positional bias mitigation=Label order reversal2026.01 | 72.2 | 87.8 |