Inference Attack on HumanEval
86.99AUCCIA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CIAMAS Framework=ARG-Designer2026.04 | 86.99 | 81.26 | 71.53 | |
| CIAMAS Framework=AGP2026.04 | 81 | 77.77 | 67.59 | |
| CIAMAS Framework=G-Designer2026.04 | 75.94 | 73.62 | 65.66 | |
| Gemini-2.5-ProMAS Framework=G-Designer2026.04 | 63.18 | 55.41 | 38.99 | |
| GPT-5MAS Framework=G-Designer2026.04 | 62.01 | 52.11 | 32.04 | |
| GPT-5MAS Framework=ARG-Designer2026.04 | 60.92 | 72.94 | 57.55 | |
| Gemini-2.5-ProMAS Framework=ARG-Designer2026.04 | 60.06 | 76.01 | 55.31 | |
| Mistral-7B-Instruct-v0.2MAS Framework=AGP2026.04 | 50 | 67.22 | 28.57 | |
| Llama-3.1-8B-InstructMAS Framework=G-Designer2026.04 | 49.22 | 46.26 | 28.63 | |
| Mistral-7B-Instruct-v0.2MAS Framework=G-Designer2026.04 | 48.93 | 44.3 | 28.11 | |
| GPT-5MAS Framework=AGP2026.04 | 44.33 | 66.61 | 26.89 | |
| Gemini-2.5-ProMAS Framework=AGP2026.04 | 40.33 | 61.22 | 21.58 | |
| Llama-3.1-8B-InstructMAS Framework=AGP2026.04 | 39.61 | 64.5 | 25 | |
| Llama-3.1-8B-InstructMAS Framework=ARG-Designer2026.04 | 34.43 | 62.04 | 34.45 | |
| Mistral-7B-Instruct-v0.2MAS Framework=ARG-Designer2026.04 | 29.36 | 58.75 | 16.53 |