Hallucination Evaluation on MMLU-Pro Law (test)
12.1HALL%GEMMA-2-9B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GEMMA-2-9Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 12.1 | 175 | 15.4 | |
| QWEN2-7Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 12.1 | 175 | 16.4 | |
| PHI-3-MINI-128Kmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 13.5 | 173 | 18 | |
| PHI-3.5-MINImode=zero-shot, extraction_framework=DecompressionLM2026.01 | 14.5 | 171 | 17.8 | |
| PHI-3-MINI-4Kmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 18.5 | 163 | 22.9 | |
| MISTRAL-NEMOmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 19.3 | 159 | 20.4 | |
| NEO-7Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 22.2 | 154 | 14.9 | |
| MINISTRAL-8Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 23.5 | 150 | 25 | |
| DEEPSEEK-MATH-7Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 25.5 | 149 | 27.9 | |
| MISTRAL-7Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 27.5 | 58 | 27.3 | |
| QWEN2-1.5Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 27.8 | 143 | 18.6 | |
| GRANITE-3.1-8Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 28 | 144 | 18.4 | |
| LLAMA-3.1-8Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 28.1 | 143 | 22.7 | |
| GRANITE-3.1-2Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 32.8 | 131 | 8.6 | |
| DEEPSEEK-V2-LITEmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 32.8 | 133 | 13.6 | |
| LLAMA-3-8Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 33.2 | 133 | -8.8 | |
| QWEN2-0.5Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 34.2 | 125 | 19.8 | |
| GRANITE-3.1-1Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 42.4 | 114 | 29.4 | |
| GRANITE-3.1-3Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 44.7 | 109 | 15.3 | |
| EXAONE-3.5-2.4Bmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 46.7 | 106 | 8.9 | |
| LLAMA-3-SMAUGmode=zero-shot, extraction_framework=DecompressionLM2026.01 | 56.8 | 86 | -14.4 |