Span-level PHI extraction on SHIELD LOCATION category 1.0 (test)
90PrecisionGPT-OSS 120B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-OSS 120Bmode=zero-shot extraction2026.05 | 90 | 61 | |
| Flash Litemode=zero-shot extraction2026.05 | 83 | 68 | |
| Gemini Flashmode=zero-shot extraction2026.05 | 78 | 59 | |
| GPT-OSS 20Bmode=zero-shot extraction2026.05 | 74 | 70 |