Retrieval-Augmented Question Answering on ClashEval
70Epsilon ObeyGemma (v3.0, 1B)
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemma (v3.0, 1B)Full Interaction Strategy=False2026.05 | 70 | |
| Gemma (v3.0, 1B)Full Interaction Strategy=True2026.05 | 57 | |
| Gemma (v3.0, 4B)Full Interaction Strategy=False2026.05 | 51 | |
| Qwen (v3.0, 4B)Full Interaction Strategy=False2026.05 | 51 | |
| Llama (v3.2, 3B)Full Interaction Strategy=False2026.05 | 50 | |
| Gemma (v3.0, 12B)Full Interaction Strategy=False2026.05 | 49 | |
| Gemma (v3.0, 4B)Full Interaction Strategy=True2026.05 | 49 | |
| Gemma (v3.0, 27B)Full Interaction Strategy=False2026.05 | 48 | |
| Llama (v3.3, 70B)Full Interaction Strategy=False2026.05 | 48 | |
| Llama (v3.2, 3B)Full Interaction Strategy=True2026.05 | 47 | |
| Qwen (v2.5, 72B)Full Interaction Strategy=False2026.05 | 46 | |
| Qwen (v3.0, 4B)Full Interaction Strategy=True2026.05 | 35 | |
| Gemma (v3.0, 12B)Full Interaction Strategy=True2026.05 | 34 | |
| Qwen (v2.5, 72B)Full Interaction Strategy=True2026.05 | 31 | |
| Llama (v3.3, 70B)Full Interaction Strategy=True2026.05 | 29 | |
| Gemma (v3.0, 27B)Full Interaction Strategy=True2026.05 | 28 |