Human-Likeness Evaluation on COGCAPTCHA30 Combined (test)
0Fool RateGemini 2.5 Pro
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Gemini 2.5 ProTraining Status=In classifier training2026.05 | 0 | 0.15 | |
| Claude SonnetTraining Status=In classifier training2026.05 | 0 | 0.02 | |
| GPT-5Training Status=In classifier training2026.05 | 4 | 0.11 | |
| Base QwenTraining Status=Held out from training, Model Size=1.5B2026.05 | 20 | 0.27 | |
| CentaurTraining Status=Held out from training, Model Size=70B2026.05 | 79 | 0.67 |