Contextual Integrity Applicability and Compliance on GoldCoin-HIPAA
96.3Applicability Macro F1Phi-4
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Phi-4Condition=Zero-shot2026.04 | 96.3 | 67 | |
| ContextReasoner-7BCondition=Zero-shot2026.04 | 96.3 | 57 | |
| CIRL-7BCondition=Zero-shot2026.04 | 95.8 | 52.5 | |
| Qwen3.5-9BCondition=Zero-shot2026.04 | 95.3 | 72.4 | |
| Gemma-3-12BCondition=Zero-shot2026.04 | 94.4 | 74.4 | |
| Qwen3.5-9BCondition=GRPO2026.04 | 93.9 | 75.3 | |
| Qwen3.5-9BCondition=SFT2026.04 | 93.4 | 73.3 | |
| Qwen3.5-4BCondition=Zero-shot2026.04 | 90.1 | 73.1 | |
| Phi-4Condition=SFT2026.04 | 90.1 | 62 | |
| OpenThinker3-7BCondition=SFT2026.04 | 83 | 15.8 | |
| Gemma-3-12BCondition=SFT2026.04 | 80.1 | 74.4 | |
| Qwen3.5-2BCondition=SFT2026.04 | 79.3 | 44.8 | |
| GPT-OSS-20BCondition=Zero-shot2026.04 | 78.6 | 73.4 | |
| OpenThinker3-7BCondition=Zero-shot2026.04 | 77.6 | 15.8 | |
| Qwen3.5-4BCondition=SFT2026.04 | 69.6 | 72.3 | |
| GPT-OSS-20BCondition=SFT2026.04 | 63.9 | 58.2 | |
| Qwen3.5-2BCondition=Zero-shot2026.04 | 56.2 | 58.6 |