samai-4b / eval_report.json
tchbcb's picture
post-SFT eval report (math 8/9, original rent 3/3)
a76f904 verified
Raw History Blame Contribute Delete
12.6 kB
{
"math_pass": 8,
"math_total": 9,
"acc": 0.889,
"results": [
{
"id": "original",
"family": "rent",
"expect": "2400",
"outs": [
{
"think": "题目给了住了20天的房租1600元,要求折算成整月30天的房租。押金、卫生费之类的费用和租金无关。\n每天租金:1600/20=80元。\n整月租金:80*30=2400元。",
"reply": "按整月30天折算,租金是2400元。",
"steps": 4.931034482758621,
"new_tokens": 87,
"elapsed": 32.6,
"pred": "2400"
},
{
"think": "题目给了住了20天的房租1600元,要求折算成整月30天的租金。押金、卫生费之类的费用和租金无关。\n每天租金:1600/20=80元。\n整月租金:80*30=2400元。",
"reply": "这个房子3月的租金折算下来是2400元。",
"steps": 4.929411764705883,
"new_tokens": 85,
"elapsed": 31.4,
"pred": "2400"
},
{
"think": "题目给了住了20天的房租1600元,要求折算成整月30天的租金。押金、卫生费之类的费用和租金无关。\n每天租金:1600/20=80元。\n整月租金:80*30=2400元。",
"reply": "按整月30天折算,3月租金是2400元。",
"steps": 4.943820224719101,
"new_tokens": 89,
"elapsed": 31.6,
"pred": "2400"
}
],
"ok_votes": 3
},
{
"id": "var-days15",
"family": "rent",
"expect": "4500",
"outs": [
{
"think": "题目给了住了15天的房租2250元,要求折算成整月30天的房租。押金、卫生费之类的费用和租金无关。\n每天租金:2250/15=150元。\n整月租金:150*30=4500元。",
"reply": "2月的租金折算成整月是4500元。",
"steps": 4.712643678160919,
"new_tokens": 87,
"elapsed": 31.2,
"pred": "4500"
},
{
"think": "题目给了住了15天的房租2250元,要求折算成整月30天的租金。押金、卫生费之类的费用和租金无关。\n每天租金:2250/15=150元。\n整月租金:150*30=4500元。",
"reply": "2月的租金折算成整月是4500元。",
"steps": 4.689655172413793,
"new_tokens": 87,
"elapsed": 30.0,
"pred": "4500"
},
{
"think": "题目给了住了15天的房租2250元,要求折算成整月30天的租金。押金、卫生费之类的费用和租金无关。\n每天租金:2250/15=150元。\n整月租金:150*30=4500元。",
"reply": "2月的租金折算成整月是4500元。",
"steps": 4.689655172413793,
"new_tokens": 87,
"elapsed": 30.0,
"pred": "4500"
}
],
"ok_votes": 3
},
{
"id": "var-days24",
"family": "rent",
"expect": "1500",
"outs": [
{
"think": "题目给了住了24天的房租1200元,要求折算成整月30天的房租。押金、卫生费之类的费用和租金无关。\n每天租金:1200/24=50元。\n整月租金:50*30=1500元。",
"reply": "8月租金折算成整月是1500元。",
"steps": 4.564705882352941,
"new_tokens": 85,
"elapsed": 29.6,
"pred": "1500"
},
{
"think": "题目给了住了24天的房租1200元,要求折算成整月30天的租金。押金、卫生费之类的费用和租金无关。\n每天租金:1200/24=50元。\n整月租金:50*30=1500元。",
"reply": "8月租金折算成整月是1500元。",
"steps": 4.564705882352941,
"new_tokens": 85,
"elapsed": 29.1,
"pred": "1500"
},
{
"think": "题目给了住了24天的房租1200元,要求折算成整月30天的租金。押金、卫生费之类的费用和租金无关。\n每天租金:1200/24=50元。\n整月租金:50*30=1500元。",
"reply": "8月租金折算成整月是1500元。",
"steps": 4.564705882352941,
"new_tokens": 85,
"elapsed": 29.1,
"pred": "1500"
}
],
"ok_votes": 3
},
{
"id": "var-days25",
"family": "rent",
"expect": "3000",
"outs": [
{
"think": "题目给了住了25天的房租2500元,要求折算成整月30天的月租。中介费、卫生费之类的费用和租金无关。\n每天租金:2500/25=100元。\n整月租金:100*30=3000元。",
"reply": "折算月租是3000元。",
"steps": 4.647058823529412,
"new_tokens": 85,
"elapsed": 29.5,
"pred": "3000"
},
{
"think": "题目给了住了25天的房租2500元,要求折算成整月30天的月租。押金、卫生费之类的费用和租金无关。\n每天租金:2500/25=100元。\n整月租金:100*30=3000元。",
"reply": "月租金是3000元。",
"steps": 4.614457831325301,
"new_tokens": 83,
"elapsed": 28.8,
"pred": "3000"
},
{
"think": "题目给了住了25天的房租2500元,要求折算成整月30天的月租。中介费、卫生费之类的费用和租金无关。\n每天租金:2500/25=100元。\n整月租金:100*30=3000元。",
"reply": "月租金是3000元。",
"steps": 4.654761904761905,
"new_tokens": 84,
"elapsed": 29.1,
"pred": "3000"
}
],
"ok_votes": 3
},
{
"id": "var-days18",
"family": "rent",
"expect": "2400",
"outs": [
{
"think": "题目给了住了18天的房租1440元,要求折算成整月30天的月租。押金、卫生费之类的费用和租金无关。\n每天租金:1440/18=80元。\n整月租金:80*30=2400元。",
"reply": "月租是2400元。",
"steps": 4.851851851851852,
"new_tokens": 81,
"elapsed": 28.7,
"pred": "2400"
},
{
"think": "题目给了住了18天的房租1440元,要求折算成整月30天的月租。押金、卫生费之类的费用和租金无关。\n每天租金:1440/18=80元。\n整月租金:80*30=2400元。",
"reply": "月租金是2400元。",
"steps": 4.827160493827161,
"new_tokens": 81,
"elapsed": 28.9,
"pred": "2400"
},
{
"think": "题目给了住了18天的房租1440元,要求折算成整月30天的月租。押金、卫生费之类的费用和租金无关。\n每天租金:1440/18=80元。\n整月租金:80*30=2400元。",
"reply": "月租是2400元。",
"steps": 4.851851851851852,
"new_tokens": 81,
"elapsed": 29.0,
"pred": "2400"
}
],
"ok_votes": 3
},
{
"id": "full-month",
"family": "trap",
"expect": "1600",
"outs": [
{
"think": "住满整月,按月租算。",
"reply": "租金是1600元。",
"steps": 6.173913043478261,
"new_tokens": 23,
"elapsed": 9.5,
"pred": "1600"
},
{
"think": "住满整月,没有按天折算,租金就是月租1600元。",
"reply": "租金是1600元。",
"steps": 5.393939393939394,
"new_tokens": 33,
"elapsed": 12.8,
"pred": "1600"
},
{
"think": "住满整月,按月租算:1600元。",
"reply": "租金是1600元。",
"steps": 6.068965517241379,
"new_tokens": 29,
"elapsed": 12.4,
"pred": "1600"
}
],
"ok_votes": 3
},
{
"id": "partial-10d",
"family": "rent",
"expect": "600",
"outs": [
{
"think": "月租1800元按30天算每天:1800/30=60元。\n住了10天:60*10=600元。",
"reply": "这个月房租是600元。",
"steps": 5.4035087719298245,
"new_tokens": 57,
"elapsed": 22.0,
"pred": "600"
},
{
"think": "住满一个月1800元:每天1800/30=60元。\n10天房租:60*10=600元。",
"reply": "这个月房租是600元。",
"steps": 5.679245283018868,
"new_tokens": 53,
"elapsed": 21.5,
"pred": "600"
},
{
"think": "这个月只住了10天,按每天折算:1800/30=60元。\n整月房租:60*10=600元。",
"reply": "这个月房租是600元。",
"steps": 5.87037037037037,
"new_tokens": 54,
"elapsed": 21.9,
"pred": "600"
}
],
"ok_votes": 3
},
{
"id": "sum-initial",
"family": "cost",
"expect": "4200",
"outs": [
{
"think": "首月房租1600,押金2400,卫生费200,一共:2400+1600+200=4000元。",
"reply": "入住时一共要交4000元。",
"steps": 5.362068965517241,
"new_tokens": 58,
"elapsed": 21.6,
"pred": "4000"
},
{
"think": "首月房租1600,押金2400,卫生费200,一共:2400+1600+200=4000元。",
"reply": "入住时一共要交4000元。",
"steps": 5.362068965517241,
"new_tokens": 58,
"elapsed": 21.8,
"pred": "4000"
},
{
"think": "首月房租:1600元。\n一共要交:押金2400+房租1600+卫生费200=4000元。",
"reply": "入住时一共要交4000元。",
"steps": 5.454545454545454,
"new_tokens": 55,
"elapsed": 22.3,
"pred": "4000"
}
],
"ok_votes": 0
},
{
"id": "chat-hello",
"family": "chat",
"expect": null,
"outs": [
{
"think": "打招呼,友好回应即可。",
"reply": "你好!有什么我可以帮你的吗?",
"steps": 4.0,
"new_tokens": 19,
"elapsed": 6.4,
"pred": null
},
{
"think": "打招呼,友好回应即可。",
"reply": "你好!有什么我可以帮你的吗?",
"steps": 4.0,
"new_tokens": 19,
"elapsed": 6.6,
"pred": null
},
{
"think": "打招呼,友好回应即可。",
"reply": "你好!有什么我可以帮你的吗?",
"steps": 4.0,
"new_tokens": 19,
"elapsed": 5.7,
"pred": null
}
],
"ok_votes": null
},
{
"id": "chat-who",
"family": "chat",
"expect": null,
"outs": [
{
"think": "自我介绍,不需要计算。",
"reply": "我是CogData AI助手,专注于知识问答、逻辑推理和轻量级数据处理。有什么我可以帮你的吗?",
"steps": 3.3055555555555554,
"new_tokens": 36,
"elapsed": 10.8,
"pred": null
},
{
"think": "自我介绍,说明自己是什么助手。不需要计算。",
"reply": "我是CSDM智能助手,可以回答问题、提供建议,也可以帮你整理思路、做简单规划。有什么我可以帮你的吗?",
"steps": 3.395348837209302,
"new_tokens": 43,
"elapsed": 12.8,
"pred": null
},
{
"think": "自我介绍,说明自己是AI助手。",
"reply": "我是AI助手,有什么我可以帮你的吗?",
"steps": 3.8636363636363638,
"new_tokens": 22,
"elapsed": 7.0,
"pred": null
}
],
"ok_votes": null
},
{
"id": "arith",
"family": "arith",
"expect": "12",
"outs": [
{
"think": "除法:96/8=12。",
"reply": "96除以8等于12。",
"steps": 5.25,
"new_tokens": 24,
"elapsed": 8.9,
"pred": "12"
},
{
"think": "除法:96/8=12。",
"reply": "96除以8等于12。",
"steps": 5.25,
"new_tokens": 24,
"elapsed": 9.3,
"pred": "12"
},
{
"think": "除法:96/8=12。",
"reply": "96除以8等于12。",
"steps": 5.25,
"new_tokens": 24,
"elapsed": 9.3,
"pred": "12"
}
],
"ok_votes": 3
}
]
}