yeji-8b-rslora-v7 / eval_comparison_base_v5_v7.json
tellang's picture
Evaluation: Base vs v5 vs v7 comparison
48549c4 verified
{
"timestamp": "2026-01-26T00:34:57.611258",
"evaluation_version": "v3_strict",
"models": {
"Base (Qwen3-8B)": {
"repo": "Qwen/Qwen3-8B-Base",
"type": "base",
"avg": 82.41666666666667,
"results": {
"사주": {
"score": 73.5,
"dimensions": {
"expertise": 60.0,
"depth": 50.0,
"quality": 100.0,
"advice": 100.0
}
},
"타로": {
"score": 100.0,
"dimensions": {
"expertise": 100.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
},
"호로스코프": {
"score": 73.75,
"dimensions": {
"expertise": 25.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
}
}
},
"v5 (LoRA)": {
"repo": "tellang/yeji-8b-lora-v5",
"type": "adapter",
"avg": 91.41666666666667,
"results": {
"사주": {
"score": 93.0,
"dimensions": {
"expertise": 80.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
},
"타로": {
"score": 90.0,
"dimensions": {
"expertise": 100.0,
"depth": 100.0,
"quality": 100.0,
"advice": 50.0
}
},
"호로스코프": {
"score": 91.25,
"dimensions": {
"expertise": 75.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
}
}
},
"v7 (rsLoRA r=64)": {
"repo": "tellang/yeji-8b-rslora-v7",
"type": "merged",
"avg": 91.83333333333333,
"results": {
"사주": {
"score": 93.0,
"dimensions": {
"expertise": 80.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
},
"타로": {
"score": 100.0,
"dimensions": {
"expertise": 100.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
},
"호로스코프": {
"score": 82.5,
"dimensions": {
"expertise": 50.0,
"depth": 100.0,
"quality": 100.0,
"advice": 100.0
}
}
}
}
}
}