Verification on VerifyBench 1.0 (test)
96.6m@3 AccuracyCoSineVerifier-Tool-4B
Evaluation Results
| Method | Links | |
|---|---|---|
| CoSineVerifier-Tool-4BVerifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=95.32025.12 | 96.6 | |
| CompassVerifier-32BVerifier Type=Labeling Verifier, Avg. Tokens=12025.12 | 96.3 | |
| o3Verifier Type=CoT Verifier, Model Access Category=Closed-source, Avg. Tokens=206.72025.12 | 96.1 | |
| GPT-4oVerifier Type=CoT Verifier, Model Access Category=Closed-source, Avg. Tokens=192.42025.12 | 96 | |
| Gemini2.5-FlashVerifier Type=CoT Verifier, Model Access Category=Closed-source, Avg. Tokens=1932025.12 | 96 | |
| CompassVerifier-32B (CoT)Verifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=2132025.12 | 95.9 | |
| CoSineVerifier-4BVerifier Type=Labeling Verifier, Avg. Tokens=12025.12 | 95.7 | |
| CoSineVerifier-32BVerifier Type=Labeling Verifier, Avg. Tokens=12025.12 | 95.7 | |
| LLaMA3.3-70B-InstructVerifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=347.32025.12 | 94.8 | |
| Qwen3-32BVerifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=798.82025.12 | 94.7 | |
| Qwen3-4B-Instruct-2507Verifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=869.72025.12 | 94.7 | |
| Qwen3-235B-A22B-2507Verifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=1885.32025.12 | 94.4 | |
| Qwen3-8BVerifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=926.62025.12 | 93.7 | |
| CompassVerifier-7B (CoT)Verifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=234.72025.12 | 93.5 | |
| CompassVerifier-7BVerifier Type=Labeling Verifier, Avg. Tokens=12025.12 | 93.5 | |
| Qwen3-4BVerifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=1156.72025.12 | 92.6 | |
| XVerify-8B-IVerifier Type=Labeling Verifier, Avg. Tokens=12025.12 | 92.5 | |
| GPT-oss-20BVerifier Type=CoT Verifier, Model Access Category=Open-source, Avg. Tokens=2212025.12 | 92.2 |