RIT Classification on IoT Security RIT dataset Experiment D
67.72Accuracy (AC)Llama-70b
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-70bPrompting Strategy=Two-Shot2026.01 | 67.72 | 0 | 40 | 62.07 | |
| Llama-70bPrompting Strategy=One-Shot2026.01 | 64.57 | 25 | 30 | 60 | |
| Llama-70bPrompting Strategy=Zero-Shot2026.01 | 62.99 | 12.5 | 40 | 58.62 | |
| Llama-8bPrompting Strategy=One-Shot2026.01 | 58.27 | 50 | 30 | 55.86 | |
| Llama-8bPrompting Strategy=Zero-Shot2026.01 | 52.76 | 25 | 10 | 48.28 | |
| Llama-8bPrompting Strategy=Two-Shot2026.01 | 50.39 | 0 | 0 | 44.14 | |
| Gemini-2.5-ProPrompting Strategy=Zero-Shot2026.01 | 48.03 | 87.5 | 70 | 51.72 | |
| Gemini-2.5-ProPrompting Strategy=One-Shot2026.01 | 44.88 | 100 | 60 | 48.97 | |
| Gemini-2.5-ProPrompting Strategy=Two-Shot2026.01 | 44.88 | 100 | 60 | 48.97 | |
| GPT-4oPrompting Strategy=Zero-Shot2026.01 | 30.71 | 87.5 | 60 | 59.4 | |
| GPT-4oPrompting Strategy=One-Shot2026.01 | 28.35 | 75 | 60 | 54.45 | |
| GPT-4oPrompting Strategy=Two-Shot2026.01 | 28.35 | 75 | 70 | 57.78 | |
| Deepseek-r1-7bPrompting Strategy=Two-Shot2026.01 | 22.05 | 0 | 40 | 22.07 | |
| Deepseek-r1-7bPrompting Strategy=One-Shot2026.01 | 21.26 | 0 | 70 | 23.45 | |
| Deepseek-r1-7bPrompting Strategy=Zero-Shot2026.01 | 19.69 | 0 | 40 | 20 |