iomgaa/InfoRSI
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
c8d89d82f1101f62b34e643d60893306ff844553
InfoRSI/tests/fixtures/rules/diagnosis.json
T
iomgaa 852600df02 feat: 诊断（任务书 18 第一块）：诊断数学、四种提交、一次运行跑多个实验、账本与材料、flow 诊断各步 
消融默认关，关掉时解释阶段一并跳过；选诊断实验用信息价值减成本。
eval/diagnosis/research.json 为第三块配置。vocab 字段说明的出口文字随提示词一起改，见下一次提交的 prompts/ 与同文件改动（本次含在内）。

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-05 10:36:20 -04:00

1938 lines
34 KiB
JSON
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
{
"generator": "gen_diagnosis.py",
"tolerance": {
"exact": "有理数精确算出的量（成分权重、事前期望收益、ĝ₀、全做的 VOI、R̂、m′、成本、价值、净值、收益、更新后的权重）：绝对误差 1e-12",
"exactK1": "只做一个实验的 VOI 是逐段闭式积分，实现用部分格网格近似：|差| ≤ 1e-9 + 2e-4·|精确值|",
"bits": "世界比特、惊讶、分歧量是浮点：绝对误差 1e-9；合并范围是分位点：绝对误差 1e-9",
"discrete": "是否诊断、选中的实验、确认与否定、有没有指出方向、实验的结局：必须完全相同"
},
"scenarios": [
{
"name": "opposite_lambda0",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.0,
"experiments": [
"d1"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.15,
"q5": 0.2,
"q9": 0.25
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": -0.25,
"q5": -0.2,
"q9": -0.15
}
}
}
],
"balance": 7,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.0
},
"g0": 0.0,
"weights": [
0.5,
0.5,
0.0
],
"combos": [
{
"experiments": [
"d1"
],
"voi": 0.1,
"rHat": 3,
"mPrime": 1.75,
"value": 0.0875,
"cost": 0.0,
"net": 0.0875
}
],
"diagnose": true,
"chosen": [
"d1"
],
"chosenNet": 0.0875
},
"note": "λ = 0，两个成分的范围不相交：做了就知道是哪一条。手算：做了以后 E[max(0, g)] = 0.5 × 0.2 = 0.1，不做时 max(0, E g) = max(0, 0.5×0.2 + 0.5×(−0.2)) = 0，VOI = 0.1。"
},
{
"name": "opposite_lambda01",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.15,
"q5": 0.2,
"q9": 0.25
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": -0.25,
"q5": -0.2,
"q9": -0.15
}
}
}
],
"balance": 7,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.0
},
"g0": 0.0,
"weights": [
0.45,
0.45,
0.1
],
"combos": [
{
"experiments": [
"d1"
],
"voi": 0.105,
"rHat": 3,
"mPrime": 1.75,
"value": 0.091875,
"cost": 0.0,
"net": 0.091875
}
],
"diagnose": true,
"chosen": [
"d1"
],
"chosenNet": 0.091875
},
"note": "同上，λ = 0.1，开放预留铺在两端之外。"
},
{
"name": "same_lambda0",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.0,
"experiments": [
"d1"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
}
}
}
],
"balance": 7,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.2
},
"g0": 0.2,
"weights": [
0.5,
0.5,
0.0
],
"combos": [
{
"experiments": [
"d1"
],
"voi": 0.0,
"rHat": 3,
"mPrime": 1.75,
"value": 0.0,
"cost": 0.03333333333333333,
"net": -0.03333333333333333
}
],
"diagnose": false,
"chosen": [],
"chosenNet": 0.0
},
"note": "λ = 0，两条解释一样，做不做都不改变看法：VOI = 0（手算：g 的取值范围 0 到 0.4 全在 0 以上，E[max(0, g)] = E g = 0.2）。净值 = −成本 < 0，不诊断。"
},
{
"name": "same_lambda01",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
}
}
}
],
"balance": 7,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.2
},
"g0": 0.2,
"weights": [
0.45,
0.45,
0.1
],
"combos": [
{
"experiments": [
"d1"
],
"voi": 0.0025,
"rHat": 3,
"mPrime": 1.75,
"value": 0.0021874999999999998,
"cost": 0.03333333333333333,
"net": -0.031145833333333334
}
],
"diagnose": false,
"chosen": [],
"chosenNet": 0.0
},
"note": "λ = 0.1：解释们的范围都在 0 以上，只有开放预留（铺在 −0.2 到 0.6）能给出负的结果，所以 VOI 只剩预留带来的一小部分。"
},
{
"name": "multi_all",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.02,
"q5": 0.05,
"q9": 0.08
},
"d2": {
"q1": -0.05,
"q5": 0.0,
"q9": 0.05
},
"d3": {
"q1": 0.0,
"q5": 0.1,
"q9": 0.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.0,
"q5": 0.03,
"q9": 0.06
},
"d2": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
},
"d3": {
"q1": -0.1,
"q5": -0.05,
"q9": 0.0
}
}
},
{
"id": "e3",
"author": "researcher-3",
"ranges": {
"d1": {
"q1": -0.02,
"q5": 0.04,
"q9": 0.1
},
"d2": {
"q1": -0.2,
"q5": -0.1,
"q9": 0.0
},
"d3": {
"q1": 0.05,
"q5": 0.1,
"q9": 0.15
}
}
}
],
"balance": 9,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.04,
"d2": 0.034999999999999996,
"d3": 0.052500000000000005
},
"g0": 0.052500000000000005,
"weights": [
0.3,
0.3,
0.3,
0.1
],
"combos": [
{
"experiments": [
"d1",
"d2",
"d3"
],
"voi": 0.10650259407848324,
"rHat": 4,
"mPrime": 1.875,
"value": 0.09984618194857804,
"cost": 0.00875,
"net": 0.09109618194857805
}
],
"diagnose": true,
"chosen": [
"d1",
"d2",
"d3"
],
"chosenNet": 0.09109618194857805
},
"note": "三个实验不超过 D = 4，只有全做一个组合。R̂ = ⌈8 ÷ 2⌉ = 4，m′ = 1.875，成本 = 2ĝ₀ × (3/4 − 2/3)。"
},
{
"name": "over_cap_d1",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.02,
"q5": 0.05,
"q9": 0.08
},
"d2": {
"q1": -0.05,
"q5": 0.0,
"q9": 0.05
},
"d3": {
"q1": 0.0,
"q5": 0.1,
"q9": 0.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.0,
"q5": 0.03,
"q9": 0.06
},
"d2": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
},
"d3": {
"q1": -0.1,
"q5": -0.05,
"q9": 0.0
}
}
},
{
"id": "e3",
"author": "researcher-3",
"ranges": {
"d1": {
"q1": -0.02,
"q5": 0.04,
"q9": 0.1
},
"d2": {
"q1": -0.2,
"q5": -0.1,
"q9": 0.0
},
"d3": {
"q1": 0.05,
"q5": 0.1,
"q9": 0.15
}
}
}
],
"balance": 9,
"batchCap": 2,
"maxExperiments": 1,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.04,
"d2": 0.034999999999999996,
"d3": 0.052500000000000005
},
"g0": 0.052500000000000005,
"weights": [
0.3,
0.3,
0.3,
0.1
],
"combos": [
{
"experiments": [
"d1"
],
"voi": 0.025331646896146795,
"rHat": 4,
"mPrime": 1.875,
"value": 0.02374841896513762,
"cost": 0.00875,
"net": 0.01499841896513762
},
{
"experiments": [
"d2"
],
"voi": 0.08911080457463884,
"rHat": 4,
"mPrime": 1.875,
"value": 0.08354137928872392,
"cost": 0.00875,
"net": 0.07479137928872391
},
{
"experiments": [
"d3"
],
"voi": 0.08340189036381514,
"rHat": 4,
"mPrime": 1.875,
"value": 0.0781892722160767,
"cost": 0.00875,
"net": 0.0694392722160767
}
],
"diagnose": true,
"chosen": [
"d2"
],
"chosenNet": 0.07479137928872391
},
"note": "三个实验超过 D = 1：只有三个单实验组合，没有空组合；精确值用逐段闭式积分。"
},
{
"name": "over_cap_d2",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.02,
"q5": 0.05,
"q9": 0.08
},
"d2": {
"q1": -0.05,
"q5": 0.0,
"q9": 0.05
},
"d3": {
"q1": 0.0,
"q5": 0.1,
"q9": 0.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.0,
"q5": 0.03,
"q9": 0.06
},
"d2": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
},
"d3": {
"q1": -0.1,
"q5": -0.05,
"q9": 0.0
}
}
},
{
"id": "e3",
"author": "researcher-3",
"ranges": {
"d1": {
"q1": -0.02,
"q5": 0.04,
"q9": 0.1
},
"d2": {
"q1": -0.2,
"q5": -0.1,
"q9": 0.0
},
"d3": {
"q1": 0.05,
"q5": 0.1,
"q9": 0.15
}
}
}
],
"balance": 9,
"batchCap": 2,
"maxExperiments": 2,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.04,
"d2": 0.034999999999999996,
"d3": 0.052500000000000005
},
"g0": 0.052500000000000005,
"weights": [
0.3,
0.3,
0.3,
0.1
],
"combos": [
{
"experiments": [
"d1",
"d2"
],
"voi": null
},
{
"experiments": [
"d1",
"d3"
],
"voi": null
},
{
"experiments": [
"d2",
"d3"
],
"voi": null
}
]
},
"note": "三个实验超过 D = 2：只有三个两实验组合，没有单实验组合；只检查组合列表（网格近似没有精确值）。"
},
{
"name": "balance_one",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.02,
"q5": 0.05,
"q9": 0.08
},
"d2": {
"q1": -0.05,
"q5": 0.0,
"q9": 0.05
},
"d3": {
"q1": 0.0,
"q5": 0.1,
"q9": 0.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.0,
"q5": 0.03,
"q9": 0.06
},
"d2": {
"q1": 0.1,
"q5": 0.2,
"q9": 0.3
},
"d3": {
"q1": -0.1,
"q5": -0.05,
"q9": 0.0
}
}
},
{
"id": "e3",
"author": "researcher-3",
"ranges": {
"d1": {
"q1": -0.02,
"q5": 0.04,
"q9": 0.1
},
"d2": {
"q1": -0.2,
"q5": -0.1,
"q9": 0.0
},
"d3": {
"q1": 0.05,
"q5": 0.1,
"q9": 0.15
}
}
}
],
"balance": 1,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.04,
"d2": 0.034999999999999996,
"d3": 0.052500000000000005
},
"g0": 0.052500000000000005,
"weights": [
0.3,
0.3,
0.3,
0.1
],
"combos": [],
"diagnose": false,
"chosen": [],
"chosenNet": 0.0
},
"note": "余额 1：没有可选组合，不诊断。"
},
{
"name": "tie_by_id",
"section": "6.12",
"input": {
"b": 0.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d10",
"d2"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d2": {
"q1": -0.2,
"q5": 0.0,
"q9": 0.2
},
"d10": {
"q1": -0.2,
"q5": 0.0,
"q9": 0.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d2": {
"q1": 0.1,
"q5": 0.3,
"q9": 0.5
},
"d10": {
"q1": 0.1,
"q5": 0.3,
"q9": 0.5
}
}
}
],
"balance": 7,
"batchCap": 2,
"maxExperiments": 1,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d2": 0.15,
"d10": 0.15
},
"g0": 0.15,
"weights": [
0.45,
0.45,
0.1
],
"combos": [
{
"experiments": [
"d2"
],
"voi": 0.07180975286849074,
"rHat": 3,
"mPrime": 1.75,
"value": 0.0628335337599294,
"cost": 0.025,
"net": 0.03783353375992939
},
{
"experiments": [
"d10"
],
"voi": 0.07180975286849074,
"rHat": 3,
"mPrime": 1.75,
"value": 0.0628335337599294,
"cost": 0.025,
"net": 0.03783353375992939
}
],
"diagnose": true,
"chosen": [
"d2"
],
"chosenNet": 0.03783353375992939
},
"note": "两个实验的预测完全一样，净值相同（差不到 1e-12），取实验编号小的 d2。"
},
{
"name": "settle_mixed",
"section": "6.12 结算；6.9",
"input": {
"b": 1.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 1.02,
"q5": 1.05,
"q9": 1.08
},
"d2": {
"q1": 0.95,
"q5": 1.0,
"q9": 1.05
},
"d3": {
"q1": 1.0,
"q5": 1.1,
"q9": 1.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 1.0,
"q5": 1.03,
"q9": 1.06
},
"d2": {
"q1": 1.1,
"q5": 1.2,
"q9": 1.3
},
"d3": {
"q1": 0.9,
"q5": 0.95,
"q9": 1.0
}
}
},
{
"id": "e3",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.98,
"q5": 1.04,
"q9": 1.1
},
"d2": {
"q1": 0.8,
"q5": 0.9,
"q9": 1.0
},
"d3": {
"q1": 1.05,
"q5": 1.1,
"q9": 1.15
}
}
}
],
"epsilon": 0.01,
"outcomes": {
"d1": {
"kind": "valid",
"score": 1.045
},
"d2": {
"kind": "insufficient"
},
"d3": {
"kind": "untested"
}
}
},
"expected": {
"experiments": {
"d1": {
"outcome": "valid",
"merged": {
"q1": 0.992,
"q5": 1.0428015564202335,
"q9": 1.0922077922077922
},
"gain": 0.04499999999999993,
"worldBits": 10.862272556463013,
"surprise": {
"status": "ok",
"bin": 450,
"p": 0.0010708333333332145,
"surprise": 9.867050331076888
},
"disagreement": 0.14778939669397495
},
"d2": {
"outcome": "insufficient",
"merged": {
"q1": 0.8466911764705882,
"q5": 1.0063258232235701,
"q9": 1.2533088235294119
}
},
"d3": {
"outcome": "untested",
"merged": {
"q1": 0.9198823529411765,
"q5": 1.068937125748503,
"q9": 1.192
}
}
},
"judgments": {
"e1": {
"d1": "confirmed"
},
"e2": {
"d1": "confirmed"
},
"e3": {
"d1": "confirmed"
}
},
"direction": {
"weights": [
0.5603112840466926,
0.2801556420233463,
0.14007782101167315,
0.019455252918287938
],
"weightsKept": false,
"gains": {
"d1": {
"gain": 0.04499999999999993,
"measured": true
},
"d2": {
"gain": 0.042996108949416356,
"measured": false
},
"d3": {
"gain": 0.05749027237354086,
"measured": false
}
},
"experiment": "d3",
"gain": 0.05749027237354086,
"found": true
}
},
"note": "d1 有效，d2 证据不足（不判、不记账），d3 没做。研究者 2 交了两条解释，各占他那份的一半。"
},
{
"name": "settle_two_valid",
"section": "6.12 结算；6.9",
"input": {
"b": 1.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 1.02,
"q5": 1.05,
"q9": 1.08
},
"d2": {
"q1": 0.95,
"q5": 1.0,
"q9": 1.05
},
"d3": {
"q1": 1.0,
"q5": 1.1,
"q9": 1.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 1.0,
"q5": 1.03,
"q9": 1.06
},
"d2": {
"q1": 1.1,
"q5": 1.2,
"q9": 1.3
},
"d3": {
"q1": 0.9,
"q5": 0.95,
"q9": 1.0
}
}
},
{
"id": "e3",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.98,
"q5": 1.04,
"q9": 1.1
},
"d2": {
"q1": 0.8,
"q5": 0.9,
"q9": 1.0
},
"d3": {
"q1": 1.05,
"q5": 1.1,
"q9": 1.15
}
}
}
],
"epsilon": 0.01,
"outcomes": {
"d1": {
"kind": "valid",
"score": 1.07
},
"d2": {
"kind": "valid",
"score": 1.25
},
"d3": {
"kind": "untested"
}
}
},
"expected": {
"experiments": {
"d1": {
"outcome": "valid",
"merged": {
"q1": 0.992,
"q5": 1.0428015564202335,
"q9": 1.0922077922077922
},
"gain": 0.07000000000000006,
"worldBits": 10.862272556463013,
"surprise": {
"status": "ok",
"bin": 700,
"p": 0.0008458333333332394,
"surprise": 10.20733896308559
},
"disagreement": 0.14778939669397495
},
"d2": {
"outcome": "valid",
"merged": {
"q1": 0.8466911764705882,
"q5": 1.0063258232235701,
"q9": 1.2533088235294119
},
"gain": 0.25,
"worldBits": 12.5514004191889,
"surprise": {
"status": "ok",
"bin": 2500,
"p": 9.714285714284636e-05,
"surprise": 13.329532555244237
},
"disagreement": 0.5374126633995484
},
"d3": {
"outcome": "untested",
"merged": {
"q1": 0.9198823529411765,
"q5": 1.068937125748503,
"q9": 1.192
}
}
},
"judgments": {
"e1": {
"d1": "confirmed",
"d2": "denied"
},
"e2": {
"d1": "denied",
"d2": "confirmed"
},
"e3": {
"d1": "confirmed",
"d2": "denied"
}
},
"direction": {
"weights": [
0.0,
0.9527410207939508,
0.0,
0.04725897920604919
],
"weightsKept": false,
"gains": {
"d1": {
"gain": 0.07000000000000006,
"measured": true
},
"d2": {
"gain": 0.25,
"measured": true
},
"d3": {
"gain": -0.04409262759924386,
"measured": false
}
},
"experiment": "d2",
"gain": 0.25,
"found": true
}
},
"note": "两个有效实验：权重按两处密度连乘更新；d2 的实测 +0.25 落在 e2 的范围里、其余两条之外。"
},
{
"name": "settle_no_direction",
"section": "6.12 结算；6.9",
"input": {
"b": 1.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 1.02,
"q5": 1.05,
"q9": 1.08
},
"d2": {
"q1": 0.95,
"q5": 1.0,
"q9": 1.05
},
"d3": {
"q1": 1.0,
"q5": 1.1,
"q9": 1.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 1.0,
"q5": 1.03,
"q9": 1.06
},
"d2": {
"q1": 1.1,
"q5": 1.2,
"q9": 1.3
},
"d3": {
"q1": 0.9,
"q5": 0.95,
"q9": 1.0
}
}
},
{
"id": "e3",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.98,
"q5": 1.04,
"q9": 1.1
},
"d2": {
"q1": 0.8,
"q5": 0.9,
"q9": 1.0
},
"d3": {
"q1": 1.05,
"q5": 1.1,
"q9": 1.15
}
}
}
],
"epsilon": 0.01,
"outcomes": {
"d1": {
"kind": "valid",
"score": 0.97
},
"d2": {
"kind": "valid",
"score": 0.8
},
"d3": {
"kind": "valid",
"score": 0.9
}
}
},
"expected": {
"experiments": {
"d1": {
"outcome": "valid",
"merged": {
"q1": 0.992,
"q5": 1.0428015564202335,
"q9": 1.0922077922077922
},
"gain": -0.030000000000000027,
"worldBits": 10.862272556463013,
"surprise": {
"status": "ok",
"bin": -300,
"p": 0.00013333333333331854,
"surprise": 12.872674880270766
},
"disagreement": 0.14778939669397495
},
"d2": {
"outcome": "valid",
"merged": {
"q1": 0.8466911764705882,
"q5": 1.0063258232235701,
"q9": 1.2533088235294119
},
"gain": -0.19999999999999996,
"worldBits": 12.5514004191889,
"surprise": {
"status": "ok",
"bin": -2000,
"p": 9.714285714284646e-05,
"surprise": 13.329532555244235
},
"disagreement": 0.5374126633995484
},
"d3": {
"outcome": "valid",
"merged": {
"q1": 0.9198823529411765,
"q5": 1.068937125748503,
"q9": 1.192
},
"gain": -0.09999999999999998,
"worldBits": 12.165720869087144,
"surprise": {
"status": "ok",
"bin": -1000,
"p": 0.00023611111111108532,
"surprise": 12.048246444854218
},
"disagreement": 0.19153258930906247
}
},
"judgments": {
"e1": {
"d1": "denied",
"d2": "denied",
"d3": "denied"
},
"e2": {
"d1": "denied",
"d2": "denied",
"d3": "confirmed"
},
"e3": {
"d1": "denied",
"d2": "confirmed",
"d3": "denied"
}
},
"direction": {
"weights": [
0.0,
0.0,
0.0,
1.0
],
"weightsKept": false,
"gains": {
"d1": {
"gain": -0.030000000000000027,
"measured": true
},
"d2": {
"gain": -0.19999999999999996,
"measured": true
},
"d3": {
"gain": -0.09999999999999998,
"measured": true
}
},
"experiment": "d1",
"gain": -0.030000000000000027,
"found": false
}
},
"note": "全部实验的收益都不为正（−0.03、−0.2、−0.1）：最大收益 −0.03 ≤ 0，记没有指出能涨分的方向。"
},
{
"name": "settle_all_insufficient",
"section": "6.12 结算；6.9",
"input": {
"b": 1.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 1.02,
"q5": 1.05,
"q9": 1.08
},
"d2": {
"q1": 0.95,
"q5": 1.0,
"q9": 1.05
},
"d3": {
"q1": 1.0,
"q5": 1.1,
"q9": 1.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 1.0,
"q5": 1.03,
"q9": 1.06
},
"d2": {
"q1": 1.1,
"q5": 1.2,
"q9": 1.3
},
"d3": {
"q1": 0.9,
"q5": 0.95,
"q9": 1.0
}
}
},
{
"id": "e3",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.98,
"q5": 1.04,
"q9": 1.1
},
"d2": {
"q1": 0.8,
"q5": 0.9,
"q9": 1.0
},
"d3": {
"q1": 1.05,
"q5": 1.1,
"q9": 1.15
}
}
}
],
"epsilon": 0.01,
"outcomes": {
"d1": {
"kind": "insufficient"
},
"d2": {
"kind": "insufficient"
},
"d3": {
"kind": "insufficient"
}
}
},
"expected": {
"experiments": {
"d1": {
"outcome": "insufficient",
"merged": {
"q1": 0.992,
"q5": 1.0428015564202335,
"q9": 1.0922077922077922
}
},
"d2": {
"outcome": "insufficient",
"merged": {
"q1": 0.8466911764705882,
"q5": 1.0063258232235701,
"q9": 1.2533088235294119
}
},
"d3": {
"outcome": "insufficient",
"merged": {
"q1": 0.9198823529411765,
"q5": 1.068937125748503,
"q9": 1.192
}
}
},
"judgments": {
"e1": {},
"e2": {},
"e3": {}
},
"direction": {
"weights": [
0.45,
0.225,
0.225,
0.1
],
"weightsKept": false,
"gains": {
"d1": {
"gain": 0.04225000000000004,
"measured": false
},
"d2": {
"gain": 0.02750000000000002,
"measured": false
},
"d3": {
"gain": 0.06375,
"measured": false
}
},
"experiment": "d3",
"gain": 0.06375,
"found": true
}
},
"note": "没有一个有效实验：不判、不记账，权重保持起始权重，方向按起始权重估的期望收益。"
},
{
"name": "settle_outside_all",
"section": "6.12 结算；6.9",
"input": {
"b": 1.0,
"direction": "maximize",
"lambda": 0.1,
"experiments": [
"d1",
"d2",
"d3"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 1.02,
"q5": 1.05,
"q9": 1.08
},
"d2": {
"q1": 0.95,
"q5": 1.0,
"q9": 1.05
},
"d3": {
"q1": 1.0,
"q5": 1.1,
"q9": 1.2
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 1.0,
"q5": 1.03,
"q9": 1.06
},
"d2": {
"q1": 1.1,
"q5": 1.2,
"q9": 1.3
},
"d3": {
"q1": 0.9,
"q5": 0.95,
"q9": 1.0
}
}
},
{
"id": "e3",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.98,
"q5": 1.04,
"q9": 1.1
},
"d2": {
"q1": 0.8,
"q5": 0.9,
"q9": 1.0
},
"d3": {
"q1": 1.05,
"q5": 1.1,
"q9": 1.15
}
}
}
],
"epsilon": 0.01,
"outcomes": {
"d1": {
"kind": "valid",
"score": 3.0
},
"d2": {
"kind": "untested"
},
"d3": {
"kind": "untested"
}
}
},
"expected": {
"experiments": {
"d1": {
"outcome": "valid",
"merged": {
"q1": 0.992,
"q5": 1.0428015564202335,
"q9": 1.0922077922077922
},
"gain": 2.0,
"worldBits": 10.862272556463013,
"surprise": {
"status": "priorZero",
"bin": 20000,
"p": 0.0
},
"disagreement": 0.14778939669397495
},
"d2": {
"outcome": "untested",
"merged": {
"q1": 0.8466911764705882,
"q5": 1.0063258232235701,
"q9": 1.2533088235294119
}
},
"d3": {
"outcome": "untested",
"merged": {
"q1": 0.9198823529411765,
"q5": 1.068937125748503,
"q9": 1.192
}
}
},
"judgments": {
"e1": {
"d1": "denied"
},
"e2": {
"d1": "denied"
},
"e3": {
"d1": "denied"
}
},
"direction": {
"weights": [
0.45,
0.225,
0.225,
0.1
],
"weightsKept": true,
"gains": {
"d1": {
"gain": 2.0,
"measured": true
},
"d2": {
"gain": 0.02750000000000002,
"measured": false
},
"d3": {
"gain": 0.06375,
"measured": false
}
},
"experiment": "d1",
"gain": 2.0,
"found": true
}
},
"note": "d1 实测 +2.0 落在所有成分（含开放预留）的范围之外，权重无从更新：保持起始权重（weightsKept）。"
},
{
"name": "minimize_all",
"section": "6.12",
"input": {
"b": 1.0,
"direction": "minimize",
"lambda": 0.1,
"experiments": [
"d1",
"d2"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.85,
"q5": 0.9,
"q9": 0.95
},
"d2": {
"q1": 0.9,
"q5": 1.0,
"q9": 1.1
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.8,
"q5": 0.85,
"q9": 0.9
},
"d2": {
"q1": 0.95,
"q5": 1.05,
"q9": 1.15
}
}
}
],
"balance": 7,
"batchCap": 2,
"maxExperiments": 4,
"u": 0.5,
"decay": 0.5
},
"expected": {
"priorGains": {
"d1": 0.125,
"d2": -0.024999999999999994
},
"g0": 0.125,
"weights": [
0.45,
0.45,
0.1
],
"combos": [
{
"experiments": [
"d1",
"d2"
],
"voi": 0.010397222222222214,
"rHat": 3,
"mPrime": 1.75,
"value": 0.009097569444444436,
"cost": 0.020833333333333332,
"net": -0.011735763888888896
}
],
"diagnose": false,
"chosen": [],
"chosenNet": 0.0
},
"note": "越小越好：收益 = b − x。d1 两条解释都预测降到 0.9 左右（收益约 +0.1），d2 预测上升。"
},
{
"name": "settle_minimize",
"section": "6.12 结算；6.9",
"input": {
"b": 1.0,
"direction": "minimize",
"lambda": 0.1,
"experiments": [
"d1",
"d2"
],
"explanations": [
{
"id": "e1",
"author": "researcher-1",
"ranges": {
"d1": {
"q1": 0.85,
"q5": 0.9,
"q9": 0.95
},
"d2": {
"q1": 0.9,
"q5": 1.0,
"q9": 1.1
}
}
},
{
"id": "e2",
"author": "researcher-2",
"ranges": {
"d1": {
"q1": 0.8,
"q5": 0.85,
"q9": 0.9
},
"d2": {
"q1": 0.95,
"q5": 1.05,
"q9": 1.15
}
}
}
],
"epsilon": 0.01,
"outcomes": {
"d1": {
"kind": "valid",
"score": 0.88
},
"d2": {
"kind": "valid",
"score": 1.05
}
}
},
"expected": {
"experiments": {
"d1": {
"outcome": "valid",
"merged": {
"q1": 0.8042553191489362,
"q5": 0.875,
"q9": 0.9457446808510638
},
"gain": 0.12,
"worldBits": 11.270948490815508,
"surprise": {
"status": "ok",
"bin": -1200,
"p": 0.0007399999999999179,
"surprise": 10.400187108808021
},
"disagreement": 0.2390359525563106
},
"d2": {
"outcome": "valid",
"merged": {
"q1": 0.8964285714285714,
"q5": 1.025,
"q9": 1.1535714285714285
},
"gain": -0.050000000000000044,
"worldBits": 12.14559114140212,
"surprise": {
"status": "ok",
"bin": 500,
"p": 0.0003711111111110703,
"surprise": 11.39586118340523
},
"disagreement": 0.11951797627815353
}
},
"judgments": {
"e1": {
"d1": "confirmed",
"d2": "confirmed"
},
"e2": {
"d1": "confirmed",
"d2": "confirmed"
}
},
"direction": {
"weights": [
0.4961715160796319,
0.496171516079633,
0.0076569678407350855
],
"weightsKept": false,
"gains": {
"d1": {
"gain": 0.12,
"measured": true
},
"d2": {
"gain": -0.050000000000000044,
"measured": true
}
},
"experiment": "d1",
"gain": 0.12,
"found": true
}
},
"note": "越小越好：d1 实测 0.88，收益 = 1.0 − 0.88 = 0.12；d2 实测 1.05，收益 −0.05。"
}
]
}
Reference in New Issue View Git Blame Copy Permalink