CAPTCHA Solving on yandex/kaleidoscope CAPTCHA zero-shot
47Solve RateHalligan
Evaluation Results
| Method | Links | |
|---|---|---|
| HalliganZero-shot evaluation protocol=true, Training status=task-specific framework2026.03 | 47 | |
| ReCAP-8BZero-shot evaluation protocol=true, Model parameter scale=8B, Training status=finetuned2026.03 | 12 | |
| ReCAP-32BZero-shot evaluation protocol=true, Model parameter scale=32B, Training status=finetuned2026.03 | 11 | |
| Qwen-8BZero-shot evaluation protocol=true, Model parameter scale=8B, Training status=base2026.03 | 4 | |
| Qwen-32BZero-shot evaluation protocol=true, Model parameter scale=32B, Training status=base2026.03 | 2 |