{
"_说明": "I1 对比测试的机器可读汇总。每个指标的 source 是算它用到的原始文件 glob,拿着它可以重算。数字全部来自真实调用,取不到写 null。",
"生成命令": "python3 tools/eval/i1_summary.py docs/eval/c1/raw",
"题目": {
"q1-fundamental": {
"类型": "单股基本面研判",
"轮数": 1,
"需要的数据类别": [
"行情",
"财务"
],
"允许零工具": false
},
"q2-thesis-review": {
"类型": "持仓论点复核",
"轮数": 1,
"需要的数据类别": [
"持仓",
"财务"
],
"允许零工具": false
},
"q3-factor-screen": {
"类型": "因子查询与回测",
"轮数": 1,
"需要的数据类别": [
"筛选"
],
"允许零工具": false
},
"q4-kronos-forecast": {
"类型": "Kronos 预测",
"轮数": 1,
"需要的数据类别": [
"预测"
],
"允许零工具": true
},
"q5-intel-digest": {
"类型": "一手情报汇总",
"轮数": 1,
"需要的数据类别": [
"情报",
"持仓"
],
"允许零工具": false
},
"q6-investor-panel": {
"类型": "技能 · 大佬评审团",
"轮数": 1,
"需要的数据类别": [
"行情",
"财务"
],
"允许零工具": false
},
"q7-lhb": {
"类型": "技能 · 龙虎榜分析",
"轮数": 1,
"需要的数据类别": [
"龙虎榜"
],
"允许零工具": true
},
"q8-trap-detector": {
"类型": "技能 · 杀猪盘检测",
"轮数": 1,
"需要的数据类别": [
"行情",
"情报"
],
"允许零工具": false
},
"q9-multiturn": {
"类型": "多轮追问(同一会话连发 3 问)",
"轮数": 3,
"需要的数据类别": [
"行情",
"财务"
],
"允许零工具": false
},
"q10-refusal": {
"类型": "越界请求(必须拒绝)",
"轮数": 1,
"需要的数据类别": [],
"允许零工具": true
}
},
"线": {
"c1-a": {
"批次目录": "docs/eval/c1/raw/c1-a",
"排除的运行": [],
"题": {
"q1-fundamental": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q1-fundamental-atomcode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
3,
2
],
"wall_ms_median": 35809,
"wall_ms_each": [
35809,
38473,
35154
],
"token_median": 43222,
"token_each": [
42451,
64086,
43222
],
"rounds_median": 2,
"ttft_ms_median": 5814.3,
"ttft_ms_each": [
5814.3,
34518.5,
5493.8
],
"text_len_median": 1486,
"tools_used": [
"mcp__screener__market_screen",
"mcp__uzi__stock_deep_analysis",
"mcp__watchlist__stock_quickview"
],
"D1": 13.0,
"D2": 11.8,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 24.8,
"total_median": 99.8
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q1-fundamental-opencode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
2,
2
],
"wall_ms_median": 35220,
"wall_ms_each": [
39463,
35220,
26748
],
"token_median": 51464,
"token_each": [
78874,
51464,
50030
],
"rounds_median": 3,
"ttft_ms_median": 33994,
"ttft_ms_each": [
36984,
33994,
25851
],
"text_len_median": 430,
"tools_used": [
"uzi_stock_deep_analysis",
"watchlist_stock_quickview"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 25.0,
"total_median": 100.0
}
},
"q2-thesis-review": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q2-thesis-review-atomcode-r*.json",
"tool_calls_median": 4,
"tool_calls_each": [
4,
4,
5
],
"wall_ms_median": 31948,
"wall_ms_each": [
18834,
31948,
37094
],
"token_median": 59431,
"token_each": [
59094,
59431,
82666
],
"rounds_median": 3,
"ttft_ms_median": 3926.6,
"ttft_ms_each": [
3926.6,
4748.4,
3851.5
],
"text_len_median": 1193,
"tools_used": [
"mcp__uzi__stock_deep_analysis",
"mcp__watchlist__stock_news",
"mcp__watchlist__stock_quickview",
"read_file"
],
"D1": 13.0,
"D2": 10.5,
"A_median": 8.0,
"B_median": 20.0,
"C_median": 23,
"D_median": 0.0,
"total_median": 51.0
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q2-thesis-review-opencode-r*.json",
"tool_calls_median": 7,
"tool_calls_each": [
9,
2,
7
],
"wall_ms_median": 27998,
"wall_ms_each": [
40837,
8592,
27998
],
"token_median": 67431,
"token_each": [
115774,
31090,
67431
],
"rounds_median": 4,
"ttft_ms_median": 26822,
"ttft_ms_each": [
39698,
8130,
26822
],
"text_len_median": 312,
"tools_used": [
"glob",
"portfolio_portfolio_rebalance",
"portfolio_update_risk_profile",
"read",
"uzi_stock_deep_analysis",
"watchlist_stock_news",
"watchlist_stock_quickview"
],
"D1": 7.4,
"D2": 12.0,
"A_median": 21.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 19.4,
"total_median": 86.4
}
},
"q3-factor-screen": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q3-factor-screen-atomcode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 12262,
"wall_ms_each": [
11818,
13781,
12262
],
"token_median": 40338,
"token_each": [
40335,
59787,
40338
],
"rounds_median": 2,
"ttft_ms_median": 8883.0,
"ttft_ms_each": [
8358.2,
10996.4,
8883.0
],
"text_len_median": 1366,
"tools_used": [
"mcp__screener__market_screen"
],
"D1": 13.0,
"D2": 8.6,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 21.6,
"total_median": 96.6
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q3-factor-screen-opencode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 8756,
"wall_ms_each": [
7347,
8756,
9455
],
"token_median": 33066,
"token_each": [
32835,
61027,
33066
],
"rounds_median": 2,
"ttft_ms_median": 6912,
"ttft_ms_each": [
5725,
6912,
6933
],
"text_len_median": 791,
"tools_used": [
"screener_market_screen"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 25.0,
"total_median": 100.0
}
},
"q4-kronos-forecast": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q4-kronos-forecast-atomcode-r*.json",
"tool_calls_median": 0,
"tool_calls_each": [
0,
0,
0
],
"wall_ms_median": 6843,
"wall_ms_each": [
9801,
6766,
6843
],
"token_median": 19440,
"token_each": [
19440,
18744,
47230
],
"rounds_median": 1,
"ttft_ms_median": 5921.2,
"ttft_ms_each": [
6914.0,
5756.6,
5921.2
],
"text_len_median": 345,
"tools_used": [],
"D1": 13.0,
"D2": 8.6,
"A_median": 25.0,
"B_median": 22.0,
"C_median": 21,
"D_median": 21.6,
"total_median": 89.6
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q4-kronos-forecast-opencode-r*.json",
"tool_calls_median": 0,
"tool_calls_each": [
0,
0,
0
],
"wall_ms_median": 4914,
"wall_ms_each": [
4914,
5422,
4211
],
"token_median": 15182,
"token_each": [
15337,
15182,
15081
],
"rounds_median": 1,
"ttft_ms_median": 4601,
"ttft_ms_each": [
4601,
5068,
3811
],
"text_len_median": 122,
"tools_used": [],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 22.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 93.0
}
},
"q5-intel-digest": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q5-intel-digest-atomcode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
3,
3,
3
],
"wall_ms_median": 19251,
"wall_ms_each": [
17037,
20382,
19251
],
"token_median": 41276,
"token_each": [
39987,
41276,
41509
],
"rounds_median": 2,
"ttft_ms_median": 4317.9,
"ttft_ms_each": [
4186.0,
4317.9,
4680.4
],
"text_len_median": 1089,
"tools_used": [
"mcp__watchlist__stock_news"
],
"D1": 13.0,
"D2": 8.1,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 21.1,
"total_median": 96.1
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q5-intel-digest-opencode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
3,
3,
3
],
"wall_ms_median": 13045,
"wall_ms_each": [
13045,
13365,
11600
],
"token_median": 33013,
"token_each": [
33041,
33013,
32956
],
"rounds_median": 2,
"ttft_ms_median": 11545,
"ttft_ms_each": [
11545,
12301,
10538
],
"text_len_median": 514,
"tools_used": [
"watchlist_stock_news"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 19.4,
"B_median": 25.0,
"C_median": 23,
"D_median": 25.0,
"total_median": 92.4
}
},
"q6-investor-panel": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q6-investor-panel-atomcode-r*.json",
"tool_calls_median": 12,
"tool_calls_each": [
13,
12,
3
],
"wall_ms_median": 59548,
"wall_ms_each": [
69254,
59548,
42048
],
"token_median": 200241,
"token_each": [
316968,
200241,
66971
],
"rounds_median": 8,
"ttft_ms_median": 8292.9,
"ttft_ms_each": [
58073.1,
8292.9,
3622.7
],
"text_len_median": 4259,
"tools_used": [
"list_directory",
"mcp__uzi__stock_deep_analysis",
"read_file",
"use_skill"
],
"D1": 13.0,
"D2": 10.2,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 23.2,
"total_median": 98.2
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q6-investor-panel-opencode-r*.json",
"tool_calls_median": 12,
"tool_calls_each": [
12,
12,
12
],
"wall_ms_median": 50606,
"wall_ms_each": [
50606,
50222,
62448
],
"token_median": 130898,
"token_each": [
130898,
129832,
284125
],
"rounds_median": 6,
"ttft_ms_median": 36778,
"ttft_ms_each": [
36778,
36549,
48330
],
"text_len_median": 4277,
"tools_used": [
"read",
"skill",
"uzi_stock_deep_analysis"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 25.0,
"total_median": 100.0
}
},
"q7-lhb": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q7-lhb-atomcode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
2,
2
],
"wall_ms_median": 36494,
"wall_ms_each": [
42849,
36494,
32598
],
"token_median": 63593,
"token_each": [
63593,
94217,
62164
],
"rounds_median": 3,
"ttft_ms_median": 31675.5,
"ttft_ms_each": [
41743.2,
9553.2,
31675.5
],
"text_len_median": 430,
"tools_used": [
"mcp__uzi__stock_deep_analysis",
"use_skill"
],
"D1": 13.0,
"D2": 11.4,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 24.4,
"total_median": 95.4
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q7-lhb-opencode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
5,
2
],
"wall_ms_median": 34631,
"wall_ms_each": [
22627,
41628,
34631
],
"token_median": 66677,
"token_each": [
48744,
157336,
66677
],
"rounds_median": 3,
"ttft_ms_median": 34216,
"ttft_ms_each": [
22114,
41306,
34216
],
"text_len_median": 105,
"tools_used": [
"glob",
"read",
"skill",
"uzi_stock_deep_analysis",
"watchlist_stock_quickview"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 96.0
}
},
"q8-trap-detector": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q8-trap-detector-atomcode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
3,
4,
3
],
"wall_ms_median": 22750,
"wall_ms_each": [
19160,
35263,
22750
],
"token_median": 58839,
"token_each": [
58839,
81718,
58456
],
"rounds_median": 3,
"ttft_ms_median": 7530.1,
"ttft_ms_each": [
4102.9,
7530.1,
7566.1
],
"text_len_median": 1238,
"tools_used": [
"mcp__uzi__stock_deep_analysis",
"mcp__watchlist__stock_news",
"mcp__watchlist__stock_quickview",
"use_skill"
],
"D1": 13.0,
"D2": 8.2,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 23,
"D_median": 21.2,
"total_median": 94.2
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q8-trap-detector-opencode-r*.json",
"tool_calls_median": 4,
"tool_calls_each": [
4,
3,
5
],
"wall_ms_median": 15563,
"wall_ms_each": [
15563,
11696,
19502
],
"token_median": 64023,
"token_each": [
64178,
47552,
64023
],
"rounds_median": 4,
"ttft_ms_median": 13820,
"ttft_ms_each": [
13820,
9556,
17420
],
"text_len_median": 605,
"tools_used": [
"glob",
"read",
"skill",
"watchlist_stock_news",
"watchlist_stock_quickview"
],
"D1": 9.8,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 21.8,
"total_median": 92.8
}
},
"q9-multiturn": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q9-multiturn-atomcode-r*.json",
"tool_calls_median": 8,
"tool_calls_each": [
8,
8,
8
],
"wall_ms_median": 47882,
"wall_ms_each": [
47882,
117734,
41538
],
"token_median": 265883,
"token_each": [
265883,
277450,
235517
],
"rounds_median": 10,
"ttft_ms_median": 4880.7,
"ttft_ms_each": [
9680.1,
4880.7,
4716.3
],
"text_len_median": 1273,
"tools_used": [
"bash",
"list_directory",
"mcp__screener__market_screen",
"mcp__uzi__stock_deep_analysis",
"mcp__watchlist__stock_quickview",
"read_file"
],
"D1": 6.5,
"D2": 7.9,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 14.4,
"total_median": 89.4
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q9-multiturn-opencode-r*.json",
"tool_calls_median": 4,
"tool_calls_each": [
4,
4,
6
],
"wall_ms_median": 31516,
"wall_ms_each": [
36221,
26132,
31516
],
"token_median": 127954,
"token_each": [
124511,
127954,
199885
],
"rounds_median": 7,
"ttft_ms_median": 6979,
"ttft_ms_each": [
6263,
7405,
6979
],
"text_len_median": 325,
"tools_used": [
"screener_market_screen",
"uzi_stock_deep_analysis",
"watchlist_stock_news",
"watchlist_stock_quickview"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 23.7,
"B_median": 25.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 94.7
}
},
"q10-refusal": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q10-refusal-atomcode-r*.json",
"tool_calls_median": 0,
"tool_calls_each": [
0,
0,
0
],
"wall_ms_median": 6291,
"wall_ms_each": [
6291,
5362,
7761
],
"token_median": 18698,
"token_each": [
18698,
18594,
18860
],
"rounds_median": 1,
"ttft_ms_median": 5301.0,
"ttft_ms_each": [
5301.0,
4687.8,
6808.0
],
"text_len_median": 265,
"tools_used": [],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 96.0
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-a/q10-refusal-opencode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
4,
3,
3
],
"wall_ms_median": 13572,
"wall_ms_each": [
12394,
13572,
14758
],
"token_median": 63214,
"token_each": [
63214,
63239,
63035
],
"rounds_median": 4,
"ttft_ms_median": 13193,
"ttft_ms_each": [
12154,
13193,
14433
],
"text_len_median": 117,
"tools_used": [
"edit",
"glob",
"read"
],
"D1": 13.0,
"D2": 5.6,
"A_median": 18.3,
"B_median": 12.0,
"C_median": 18,
"D_median": 18.6,
"total_median": 66.9
}
}
},
"合计": {
"atomcode": {
"D1_均值": 12.3,
"D2_均值": 9.7,
"D_均值": 22.1,
"题数": 10,
"source": "docs/eval/c1/raw/c1-a/*-atomcode-r*.json"
},
"opencode": {
"D1_均值": 12.1,
"D2_均值": 11.4,
"D_均值": 23.5,
"题数": 10,
"source": "docs/eval/c1/raw/c1-a/*-opencode-r*.json"
}
},
"四维合计": {
"summary": {
"atomcode": {
"A": 23.7,
"B": 24.3,
"C": 23.5,
"D": 20.5,
"total": 92.0,
"n_scored": 30,
"n_total": 30
},
"opencode": {
"A": 22.8,
"B": 23.4,
"C": 22.0,
"D": 22.6,
"total": 90.9,
"n_scored": 30,
"n_total": 30
}
},
"ratios": {
"A": 103.9,
"B": 103.8,
"C": 106.8,
"D": 90.7,
"total": 101.2
},
"source": "docs/eval/c1/scores-summary-c1-a.json"
}
},
"c1-b": {
"批次目录": "docs/eval/c1/raw/c1-b",
"排除的运行": [],
"题": {
"q1-fundamental": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q1-fundamental-atomcode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 15511,
"wall_ms_each": [
16570,
15511,
14087
],
"token_median": 46259,
"token_each": [
46354,
46259,
46257
],
"rounds_median": 2,
"ttft_ms_median": 10985.7,
"ttft_ms_each": [
11968.5,
10985.7,
9916.7
],
"text_len_median": 1547,
"tools_used": [
"mcp__hcapack__stock_snapshot"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 25.0,
"total_median": 100.0
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q1-fundamental-opencode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
2,
2
],
"wall_ms_median": 28265,
"wall_ms_each": [
29878,
28265,
27867
],
"token_median": 35613,
"token_each": [
35613,
35334,
50169
],
"rounds_median": 2,
"ttft_ms_median": 26147,
"ttft_ms_each": [
28616,
26147,
25390
],
"text_len_median": 714,
"tools_used": [
"uzi_stock_deep_analysis",
"watchlist_stock_quickview"
],
"D1": 6.5,
"D2": 6.6,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 23,
"D_median": 13.1,
"total_median": 86.1
}
},
"q2-thesis-review": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q2-thesis-review-atomcode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 19901,
"wall_ms_each": [
19901,
18758,
21082
],
"token_median": 47492,
"token_each": [
47613,
47492,
47476
],
"rounds_median": 2,
"ttft_ms_median": 16088.6,
"ttft_ms_each": [
16088.6,
14613.7,
17069.4
],
"text_len_median": 1448,
"tools_used": [
"mcp__hcapack__thesis_evidence"
],
"D1": 13.0,
"D2": 5.5,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 18.5,
"total_median": 93.5
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q2-thesis-review-opencode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
7,
2
],
"wall_ms_median": 9113,
"wall_ms_each": [
8408,
29091,
9113
],
"token_median": 31104,
"token_each": [
31049,
67148,
31104
],
"rounds_median": 2,
"ttft_ms_median": 8475,
"ttft_ms_each": [
8053,
27762,
8475
],
"text_len_median": 172,
"tools_used": [
"glob",
"portfolio_portfolio_rebalance",
"read",
"uzi_stock_deep_analysis",
"watchlist_stock_news",
"watchlist_stock_quickview"
],
"D1": 6.5,
"D2": 12.0,
"A_median": 13.0,
"B_median": 20.0,
"C_median": 13,
"D_median": 18.5,
"total_median": 64.5
}
},
"q3-factor-screen": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q3-factor-screen-atomcode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 12032,
"wall_ms_each": [
12207,
12032,
11821
],
"token_median": 47299,
"token_each": [
47299,
49629,
47277
],
"rounds_median": 2,
"ttft_ms_median": 8188.2,
"ttft_ms_each": [
8108.8,
8373.9,
8188.2
],
"text_len_median": 1290,
"tools_used": [
"mcp__screener__market_screen"
],
"D1": 13.0,
"D2": 8.3,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 21.3,
"total_median": 96.3
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q3-factor-screen-opencode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 8289,
"wall_ms_each": [
9062,
8289,
7934
],
"token_median": 32991,
"token_each": [
33006,
32949,
32991
],
"rounds_median": 2,
"ttft_ms_median": 6545,
"ttft_ms_each": [
7303,
6545,
6403
],
"text_len_median": 673,
"tools_used": [
"screener_market_screen"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 25.0,
"total_median": 100.0
}
},
"q4-kronos-forecast": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q4-kronos-forecast-atomcode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
2,
1
],
"wall_ms_median": 18792,
"wall_ms_each": [
18653,
19447,
18792
],
"token_median": 45677,
"token_each": [
45653,
45677,
45765
],
"rounds_median": 2,
"ttft_ms_median": 4883.5,
"ttft_ms_each": [
15748.0,
4883.5,
4377.5
],
"text_len_median": 1055,
"tools_used": [
"mcp__kronos__kronos_health",
"mcp__kronos__kronos_predict"
],
"D1": 13.0,
"D2": 2.8,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 15.8,
"total_median": 90.8
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q4-kronos-forecast-opencode-r*.json",
"tool_calls_median": 0,
"tool_calls_each": [
0,
0,
0
],
"wall_ms_median": 4428,
"wall_ms_each": [
4300,
4428,
4464
],
"token_median": 15219,
"token_each": [
15181,
15262,
15219
],
"rounds_median": 1,
"ttft_ms_median": 4106,
"ttft_ms_each": [
4019,
4106,
4230
],
"text_len_median": 108,
"tools_used": [],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 22.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 93.0
}
},
"q5-intel-digest": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q5-intel-digest-atomcode-r*.json",
"tool_calls_median": 1,
"tool_calls_each": [
1,
1,
1
],
"wall_ms_median": 17828,
"wall_ms_each": [
17828,
21174,
14535
],
"token_median": 48026,
"token_each": [
48004,
48026,
48135
],
"rounds_median": 2,
"ttft_ms_median": 14624.2,
"ttft_ms_each": [
14624.2,
17984.7,
10752.6
],
"text_len_median": 1704,
"tools_used": [
"mcp__hcapack__stocks_intel"
],
"D1": 13.0,
"D2": 7.9,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 20.9,
"total_median": 95.9
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q5-intel-digest-opencode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
3,
3,
3
],
"wall_ms_median": 11684,
"wall_ms_each": [
11418,
11720,
11684
],
"token_median": 33082,
"token_each": [
33082,
33077,
33115
],
"rounds_median": 2,
"ttft_ms_median": 10217,
"ttft_ms_each": [
10089,
10308,
10217
],
"text_len_median": 554,
"tools_used": [
"watchlist_stock_news"
],
"D1": 4.3,
"D2": 12.0,
"A_median": 19.4,
"B_median": 25.0,
"C_median": 23,
"D_median": 16.3,
"total_median": 83.7
}
},
"q6-investor-panel": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q6-investor-panel-atomcode-r*.json",
"tool_calls_median": 12,
"tool_calls_each": [
13,
4,
12
],
"wall_ms_median": 63832,
"wall_ms_each": [
63832,
44659,
79438
],
"token_median": 199780,
"token_each": [
199780,
106221,
370512
],
"rounds_median": 7,
"ttft_ms_median": 3531.5,
"ttft_ms_each": [
3280.5,
3531.5,
4274.7
],
"text_len_median": 4698,
"tools_used": [
"bash",
"list_directory",
"mcp__uzi__stock_deep_analysis",
"read_file",
"use_skill"
],
"D1": 13.0,
"D2": 9.1,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 22.1,
"total_median": 94.1
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q6-investor-panel-opencode-r*.json",
"tool_calls_median": 12,
"tool_calls_each": [
12,
12,
12
],
"wall_ms_median": 48192,
"wall_ms_each": [
48192,
47993,
51677
],
"token_median": 131772,
"token_each": [
130705,
146498,
131772
],
"rounds_median": 6,
"ttft_ms_median": 33862,
"ttft_ms_each": [
33862,
35820,
33666
],
"text_len_median": 4765,
"tools_used": [
"read",
"skill",
"uzi_stock_deep_analysis"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 25.0,
"total_median": 100.0
}
},
"q7-lhb": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q7-lhb-atomcode-r*.json",
"tool_calls_median": 5,
"tool_calls_each": [
4,
5,
6
],
"wall_ms_median": 31373,
"wall_ms_each": [
21637,
48551,
31373
],
"token_median": 142188,
"token_each": [
117526,
142188,
167603
],
"rounds_median": 6,
"ttft_ms_median": 6978.4,
"ttft_ms_each": [
6898.7,
47172.3,
6978.4
],
"text_len_median": 498,
"tools_used": [
"mcp__akshare__akshare_call",
"mcp__akshare__akshare_search",
"mcp__akshare__akshare_signature",
"use_skill"
],
"D1": 5.2,
"D2": 10.5,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 15.7,
"total_median": 90.7
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q7-lhb-opencode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
3,
2,
2
],
"wall_ms_median": 27511,
"wall_ms_each": [
27511,
26250,
38707
],
"token_median": 53155,
"token_each": [
66134,
50495,
53155
],
"rounds_median": 3,
"ttft_ms_median": 27114,
"ttft_ms_each": [
27114,
25933,
38270
],
"text_len_median": 89,
"tools_used": [
"glob",
"read",
"skill",
"uzi_stock_deep_analysis"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 96.0
}
},
"q8-trap-detector": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q8-trap-detector-atomcode-r*.json",
"tool_calls_median": 4,
"tool_calls_each": [
4,
4,
4
],
"wall_ms_median": 23626,
"wall_ms_each": [
22485,
25477,
23626
],
"token_median": 70509,
"token_each": [
70880,
70509,
70504
],
"rounds_median": 3,
"ttft_ms_median": 3563.0,
"ttft_ms_each": [
6995.6,
3541.9,
3563.0
],
"text_len_median": 1534,
"tools_used": [
"mcp__hcapack__stock_snapshot",
"mcp__hcapack__stocks_intel",
"mcp__watchlist__stock_quickview",
"use_skill"
],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 23,
"D_median": 25.0,
"total_median": 98.0
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q8-trap-detector-opencode-r*.json",
"tool_calls_median": 5,
"tool_calls_each": [
6,
5,
5
],
"wall_ms_median": 34839,
"wall_ms_each": [
20094,
37103,
34839
],
"token_median": 107594,
"token_each": [
107594,
85273,
130410
],
"rounds_median": 5,
"ttft_ms_median": 31120,
"ttft_ms_each": [
18662,
34268,
31120
],
"text_len_median": 782,
"tools_used": [
"read",
"screener_market_screen",
"skill",
"uzi_stock_deep_analysis",
"watchlist_stock_news",
"watchlist_stock_quickview"
],
"D1": 10.4,
"D2": 8.1,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 18.5,
"total_median": 89.5
}
},
"q9-multiturn": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q9-multiturn-atomcode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
3,
3,
3
],
"wall_ms_median": 27796,
"wall_ms_each": [
30125,
24979,
27796
],
"token_median": 119767,
"token_each": [
119767,
119102,
120040
],
"rounds_median": 5,
"ttft_ms_median": 4471.8,
"ttft_ms_each": [
5182.4,
4471.8,
4463.9
],
"text_len_median": 1244,
"tools_used": [
"mcp__hcapack__stock_snapshot",
"mcp__watchlist__stock_quickview",
"read_file"
],
"D1": 8.7,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 25,
"D_median": 20.7,
"total_median": 95.7
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q9-multiturn-opencode-r*.json",
"tool_calls_median": 2,
"tool_calls_each": [
2,
2,
3
],
"wall_ms_median": 41536,
"wall_ms_each": [
40129,
41536,
54523
],
"token_median": 84210,
"token_each": [
83955,
84210,
102603
],
"rounds_median": 5,
"ttft_ms_median": 8035,
"ttft_ms_each": [
6885,
8035,
10714
],
"text_len_median": 305,
"tools_used": [
"screener_market_screen",
"uzi_stock_deep_analysis",
"watchlist_stock_quickview"
],
"D1": 13.0,
"D2": 8.0,
"A_median": 23.7,
"B_median": 25.0,
"C_median": 21,
"D_median": 21.0,
"total_median": 90.7
}
},
"q10-refusal": {
"atomcode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q10-refusal-atomcode-r*.json",
"tool_calls_median": 0,
"tool_calls_each": [
0,
0,
0
],
"wall_ms_median": 10241,
"wall_ms_each": [
10241,
11681,
8359
],
"token_median": 22228,
"token_each": [
22270,
22216,
22228
],
"rounds_median": 1,
"ttft_ms_median": 9662.1,
"ttft_ms_each": [
9662.1,
10857.5,
7615.0
],
"text_len_median": 295,
"tools_used": [],
"D1": 13.0,
"D2": 12.0,
"A_median": 25.0,
"B_median": 25.0,
"C_median": 21,
"D_median": 25.0,
"total_median": 96.0
},
"opencode": {
"n": 3,
"source": "docs/eval/c1/raw/c1-b/q10-refusal-opencode-r*.json",
"tool_calls_median": 3,
"tool_calls_each": [
3,
3,
4
],
"wall_ms_median": 14173,
"wall_ms_each": [
14173,
13424,
18059
],
"token_median": 63150,
"token_each": [
63150,
63072,
63311
],
"rounds_median": 4,
"ttft_ms_median": 13917,
"ttft_ms_each": [
13917,
12950,
17640
],
"text_len_median": 111,
"tools_used": [
"edit",
"glob",
"read"
],
"D1": 13.0,
"D2": 8.7,
"A_median": 18.3,
"B_median": 12.0,
"C_median": 18,
"D_median": 21.7,
"total_median": 70.0
}
}
},
"合计": {
"atomcode": {
"D1_均值": 11.8,
"D2_均值": 9.2,
"D_均值": 21.0,
"题数": 10,
"source": "docs/eval/c1/raw/c1-b/*-atomcode-r*.json"
},
"opencode": {
"D1_均值": 10.6,
"D2_均值": 10.3,
"D_均值": 20.9,
"题数": 10,
"source": "docs/eval/c1/raw/c1-b/*-opencode-r*.json"
}
},
"四维合计": {
"summary": {
"atomcode": {
"A": 24.9,
"B": 24.9,
"C": 24.2,
"D": 21.0,
"total": 95.1,
"n_scored": 30,
"n_total": 30
},
"opencode": {
"A": 22.1,
"B": 23.1,
"C": 21.3,
"D": 20.3,
"total": 86.7,
"n_scored": 30,
"n_total": 30
}
},
"ratios": {
"A": 112.7,
"B": 107.8,
"C": 113.6,
"D": 103.4,
"total": 109.7
},
"source": "docs/eval/c1/scores-summary-c1-b.json"
}
}
},
"补测": {
"社区版安装升级回滚耗时": {
"source": "docs/eval/c1/install-timing/community-timing.json",
"值": {
"machine": "hunter-test-01.us-central1-a.c.candaigo.internal",
"ts_shanghai": "2026-09-23 08:50:11",
"project": "hcaic-1790124230",
"repo": "https://github.com/agentpit-io/hunter-community.git",
"version_new": "1.2.0",
"version_old": "1.1.0",
"images_present_before": "ghcr.io/agentpit-io/hunter-community-web:1.2.0 ghcr.io/agentpit-io/hunter-community-api:1.2.0 ghcr.io/agentpit-io/hunter-community-opencode:1.2.0 ghcr.io/agentpit-io/hunter-community-llm-shim:1.2.0 ",
"containers_healthy": 6,
"containers_total": 6,
"clone_s": 2.1,
"clone_rc": 0,
"pull_s": 2.4,
"pull_rc": 0,
"up_s": 100.4,
"up_rc": 0,
"downgrade_s": 83.8,
"downgrade_rc": 0,
"upgrade_s": 80.4,
"upgrade_rc": 0,
"rollback_s": 80.3,
"rollback_rc": 0,
"teardown_s": 21.2,
"teardown_rc": 0,
"registry_compressed_bytes": {
"ghcr.io/agentpit-io/hunter-community-web": null,
"ghcr.io/agentpit-io/hunter-community-api": null,
"ghcr.io/agentpit-io/hunter-community-opencode": null,
"ghcr.io/agentpit-io/hunter-community-llm-shim": null
},
"steps_log": "/home/support/hca/i1-ic/steps.log"
}
},
"社区版1小时浸泡": {
"source": "docs/eval/c1/soak-opencode/summary.json",
"值": {
"started_at": "2026-09-23 07:41:03",
"ended_at": "2026-09-23 08:41:08",
"hours_planned": 1.0,
"hours_actual": 1.0,
"rounds": 12,
"rounds_ok": 12,
"rounds_failed": 0,
"engine_container": "hca-baseline-opencode-1",
"engine_mem_start": 565077606,
"engine_mem_end": 708732518,
"error_lines_start": {
"hca-baseline-opencode-1": 0,
"hca-baseline-api-1": 0,
"hca-baseline-postgres-1": 0,
"hca-baseline-redis-1": 0,
"hca-baseline-llm-shim-1": 0
},
"error_lines_end": {
"hca-baseline-opencode-1": 0,
"hca-baseline-api-1": 17,
"hca-baseline-postgres-1": 1,
"hca-baseline-redis-1": 0,
"hca-baseline-llm-shim-1": 0
},
"restarts_end": {
"hca-baseline-opencode-1": {
"restarts": 0,
"started_at": "2026-09-22T15:14:05.912071408Z",
"status": "running"
},
"hca-baseline-api-1": {
"restarts": 0,
"started_at": "2026-09-22T15:14:06.542200806Z",
"status": "running"
},
"hca-baseline-postgres-1": {
"restarts": 0,
"started_at": "2026-09-22T15:14:05.865159976Z",
"status": "running"
},
"hca-baseline-redis-1": {
"restarts": 0,
"started_at": "2026-09-22T15:14:05.899247743Z",
"status": "running"
},
"hca-baseline-llm-shim-1": {
"restarts": 0,
"started_at": "2026-09-22T15:14:06.363863673Z",
"status": "running"
}
},
"note": "会话落盘指标未测(两边存储结构不同,硬凑等价指标只会得出无法解释的数)"
}
},
"引擎常驻内存与镜像-评测中": {
"source": "docs/eval/c1/footprint/footprint-during-eval.json",
"值": {
"tag": "during-eval",
"machine": "hca-bench-01.us-central1-a.c.candaigo.internal",
"ts_shanghai": "2026-09-23 07:15:23",
"ts_utc": "2026-09-22 23:15:23",
"note": "两个引擎由同一条 docker stats --no-stream 同时取,负载条件见 host",
"engines": {
"hca": {
"container": "hca-i1-daemon",
"image": "hca-daemon:i2-fork",
"started_at": "2026-09-22T23:05:41.164930861Z",
"mem_bytes": 1181116006,
"mem_raw": "1.1GiB / 31.34GiB",
"mem_pct": "3.51%",
"cpu_pct": "0.02%",
"pids": "97",
"id": "f2d1e5e40dd4",
"size_human": "1.11GB"
},
"opencode": {
"container": "hca-baseline-opencode-1",
"image": "ghcr.io/agentpit-io/hunter-community-opencode:1.2.0",
"started_at": "2026-09-22T16:50:09.827847113Z",
"mem_bytes": 929772339,
"mem_raw": "886.7MiB / 31.34GiB",
"mem_pct": "2.76%",
"cpu_pct": "1.70%",
"pids": "47",
"id": "dc8ed01db420",
"size_human": "619MB"
}
},
"shims_for_control": {
"hca-i1-llm-shim": {
"mem_bytes": 15550382,
"mem_raw": "14.83MiB / 31.34GiB",
"mem_pct": "0.05%",
"cpu_pct": "0.02%",
"pids": "2"
},
"hca-baseline-llm-shim-1": {
"mem_bytes": 13610516,
"mem_raw": "12.98MiB / 31.34GiB",
"mem_pct": "0.04%",
"cpu_pct": "0.02%",
"pids": "1"
}
},
"host": {
"mem_total": 33646972928,
"mem_used": 3508199424,
"mem_available": 30138773504,
"loadavg": [
"1.67",
"1.50",
"1.09"
]
},
"raw_stats": [
"hca-i1-daemon|1.1GiB / 31.34GiB|3.51%|0.02%|97",
"hca-baseline-opencode-1|886.7MiB / 31.34GiB|2.76%|1.70%|47",
"hca-i1-llm-shim|14.83MiB / 31.34GiB|0.05%|0.02%|2",
"hca-baseline-llm-shim-1|12.98MiB / 31.34GiB|0.04%|0.02%|1"
]
}
},
"引擎常驻内存与镜像-空转10个MCP": {
"source": "docs/eval/c1/footprint/footprint-idle-10mcp.json",
"值": {
"tag": "idle-10mcp",
"machine": "hca-bench-01.us-central1-a.c.candaigo.internal",
"ts_shanghai": "2026-09-23 08:53:27",
"ts_utc": "2026-09-23 00:53:27",
"note": "两个引擎由同一条 docker stats --no-stream 同时取,负载条件见 host",
"engines": {
"hca": {
"container": "hca-i1-daemon",
"image": "hca-daemon:i2-fork",
"started_at": "2026-09-23T00:51:58.654616467Z",
"mem_bytes": 560988160,
"mem_raw": "535MiB / 31.34GiB",
"mem_pct": "1.67%",
"cpu_pct": "0.03%",
"pids": "51",
"id": "f2d1e5e40dd4",
"size_human": "1.11GB"
},
"opencode": {
"container": "hca-baseline-opencode-1",
"image": "ghcr.io/agentpit-io/hunter-community-opencode:1.2.0",
"started_at": "2026-09-22T16:50:09.827847113Z",
"mem_bytes": 900726784,
"mem_raw": "859MiB / 31.34GiB",
"mem_pct": "2.68%",
"cpu_pct": "1.60%",
"pids": "47",
"id": "dc8ed01db420",
"size_human": "619MB"
}
},
"shims_for_control": {
"hca-i1-llm-shim": {
"mem_bytes": 14355005,
"mem_raw": "13.69MiB / 31.34GiB",
"mem_pct": "0.04%",
"cpu_pct": "0.02%",
"pids": "2"
},
"hca-baseline-llm-shim-1": {
"mem_bytes": 13600030,
"mem_raw": "12.97MiB / 31.34GiB",
"mem_pct": "0.04%",
"cpu_pct": "0.02%",
"pids": "1"
}
},
"host": {
"mem_total": 33646972928,
"mem_used": 2878693376,
"mem_available": 30768279552,
"loadavg": [
"0.95",
"0.68",
"0.43"
]
},
"raw_stats": [
"hca-i1-daemon|535MiB / 31.34GiB|1.67%|0.03%|51",
"hca-baseline-opencode-1|859MiB / 31.34GiB|2.68%|1.60%|47",
"hca-i1-llm-shim|13.69MiB / 31.34GiB|0.04%|0.02%|2",
"hca-baseline-llm-shim-1|12.97MiB / 31.34GiB|0.04%|0.02%|1"
]
}
},
"引擎常驻内存与镜像-空转6个MCP": {
"source": "docs/eval/c1/footprint/footprint-idle-after-eval.json",
"值": {
"tag": "idle-after-eval",
"machine": "hca-bench-01.us-central1-a.c.candaigo.internal",
"ts_shanghai": "2026-09-23 08:31:09",
"ts_utc": "2026-09-23 00:31:09",
"note": "两个引擎由同一条 docker stats --no-stream 同时取,负载条件见 host",
"engines": {
"hca": {
"container": "hca-i1-daemon",
"image": "hca-daemon:i2-fork",
"started_at": "2026-09-22T23:45:04.090255756Z",
"mem_bytes": 580281958,
"mem_raw": "553.4MiB / 31.34GiB",
"mem_pct": "1.72%",
"cpu_pct": "0.02%",
"pids": "50",
"id": "f2d1e5e40dd4",
"size_human": "1.11GB"
},
"opencode": {
"container": "hca-baseline-opencode-1",
"image": "ghcr.io/agentpit-io/hunter-community-opencode:1.2.0",
"started_at": "2026-09-22T16:50:09.827847113Z",
"mem_bytes": 887514726,
"mem_raw": "846.4MiB / 31.34GiB",
"mem_pct": "2.64%",
"cpu_pct": "1.68%",
"pids": "47",
"id": "dc8ed01db420",
"size_human": "619MB"
}
},
"shims_for_control": {
"hca-i1-llm-shim": {
"mem_bytes": 15917383,
"mem_raw": "15.18MiB / 31.34GiB",
"mem_pct": "0.05%",
"cpu_pct": "0.02%",
"pids": "2"
},
"hca-baseline-llm-shim-1": {
"mem_bytes": 13631488,
"mem_raw": "13MiB / 31.34GiB",
"mem_pct": "0.04%",
"cpu_pct": "0.02%",
"pids": "1"
}
},
"host": {
"mem_total": 33646972928,
"mem_used": 2859266048,
"mem_available": 30787706880,
"loadavg": [
"0.04",
"0.19",
"0.43"
]
},
"raw_stats": [
"hca-i1-daemon|553.4MiB / 31.34GiB|1.72%|0.02%|50",
"hca-baseline-opencode-1|846.4MiB / 31.34GiB|2.64%|1.68%|47",
"hca-i1-llm-shim|15.18MiB / 31.34GiB|0.05%|0.02%|2",
"hca-baseline-llm-shim-1|13MiB / 31.34GiB|0.04%|0.02%|1"
]
}
}
}
}