Human process distribution matching on Wisconsin Card Sorting Test (WCST)
0.48Mean Absolute Cohen's dClaude Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude Sonnet2026.05 | 0.48 | |
| CentaurParameter Scale=70B2026.05 | 0.76 | |
| Base QwenParameter Scale=1.5B2026.05 | 0.88 | |
| Gemini 2.5 Pro2026.05 | 1.38 | |
| GPT-52026.05 | 1.91 |