Abstract Screening on Review 1 821 abstracts (Final Includes)
45False PositivesClaude sonnet
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Claude sonnet2025.12 | 45 | — | 52.38 | 94 | 90.77 | 42.31 | 0.68 | 0.37 | 0.38 | |
| gpt-52025.12 | 82 | — | 90.48 | 88.56 | 88.72 | 41.01 | 0.8 | 0.36 | 0.4 | |
| grok-4-fast2025.12 | 96 | — | 92.06 | 87.2 | 87.58 | 37.66 | 0.76 | 0.53 | 0.4 | |
| gpt-4o-mini2025.12 | 149 | — | 71.43 | 80.05 | 79.38 | 23.2 | 0.55 | 0.62 | 0.67 | |
| gpt-5-mini2025.12 | 159 | — | 100 | 78.66 | 80.32 | 28.38 | 0.76 | 0.46 | 0.53 | |
| Gemini flash2025.12 | 173 | — | 98.41 | 76.56 | 78.28 | 26.38 | 0.75 | 0.63 | 0.51 | |
| Claude haiku2025.12 | 245 | — | 85.71 | 66.44 | 67.97 | 18.06 | 0.61 | 0.69 | 0.69 | |
| gpt-5-nano2025.12 | 260 | — | 98.41 | 65.24 | 67.82 | 19.25 | 0.75 | 0.42 | 0.53 |