惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Microsoft Azure Blog
Microsoft Azure Blog
WordPress大学
WordPress大学
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
The Cloudflare Blog
U
Unit 42
D
Docker
Hugging Face - Blog
Hugging Face - Blog
博客园 - 聂微东
Recent Announcements
Recent Announcements
GbyAI
GbyAI
T
The Blog of Author Tim Ferriss
Last Week in AI
Last Week in AI
V
Visual Studio Blog
I
InfoQ
Google DeepMind News
Google DeepMind News
小众软件
小众软件
L
LangChain Blog
C
Check Point Blog
宝玉的分享
宝玉的分享
Martin Fowler
Martin Fowler
酷 壳 – CoolShell
酷 壳 – CoolShell
博客园 - 【当耐特】
J
Java Code Geeks
罗磊的独立博客

Recent Commits to openclaw:main

test: merge chat side-result checks · openclaw/openclaw@ddd2c2a test: merge cron history checks · openclaw/openclaw@f7eb746 test: merge responsive navigation shell checks · openclaw/openclaw@c2e4b47 docs(changelog): add codex oauth fixes · openclaw/openclaw@628e6cd test: merge navigation routing cases · openclaw/openclaw@5d8cecb Tests: mock channel registry bundled fallback · openclaw/openclaw@2b08233 Secrets: avoid broad web search discovery for single plugin config · openclaw/openclaw@a464f59 test: merge config view browser checks · openclaw/openclaw@20cf511 fix(status): align oauth health with runtime · openclaw/openclaw@eed7116 feat: add macOS screen snapshots for monitor preview (#67954) thanks … · openclaw/openclaw@f377db1 fix: report shared auth scopes in hello-ok (#67810) thanks @BunsDev · openclaw/openclaw@0b6c39b Auto-reply: avoid eager bundled route fallback · openclaw/openclaw@3ea1bf4 Tests: narrow session binding contract setup · openclaw/openclaw@54e4e16 fix(macOS): enable undo/redo in webchat composer text input (#34962) · openclaw/openclaw@00951dc Tests: speed up channel setup promotion · openclaw/openclaw@82b529a Docs: refresh agent instructions · openclaw/openclaw@5775fe2 fix(auth): serialize OAuth refresh across agents to fix #26322 (#67876) · openclaw/openclaw@8e79080 test: allow ollama public surface boundary test · openclaw/openclaw@7d4f1a6 Docs: add test performance guardrails · openclaw/openclaw@89706d3 Tests: restore context-engine usage proof · openclaw/openclaw@e4c4f95 Tests: slim context engine runtime coverage · openclaw/openclaw@74c198f ci: retry failed custom checkouts · openclaw/openclaw@0ee5baf test: trim duplicate provider auth onboarding cases · openclaw/openclaw@1ffc02e matrix: fix sessions_spawn --thread subagent session spawning (#67643) · openclaw/openclaw@1ce2596 test: reduce auth choice fixture churn · openclaw/openclaw@857b9cd test: mock health status config boundaries · openclaw/openclaw@9d5ab4a test: mock onboard config io boundary · openclaw/openclaw@299694d test: mock legacy state plugin boundaries · openclaw/openclaw@2713089 test: mock channel install boundaries · openclaw/openclaw@b945248 test: mock doctor preview channel boundaries · openclaw/openclaw@b1a3ad4
feat(dreaming): score candidates with shadow trial result...
iFiras-Max1 · 2026-06-01 · via Recent Commits to openclaw:main

@@ -36,6 +36,30 @@ export type DreamingShadowTrialReport = {

3636

markdown: string;

3737

};

383839+

export type DreamingShadowTrialScoreOptions = {

40+

helpfulBoost?: number;

41+

neutralDelta?: number;

42+

harmfulPenalty?: number;

43+

};

44+45+

export type DreamingShadowTrialCandidateInput = {

46+

key: string;

47+

score: number;

48+

};

49+50+

export type DreamingShadowTrialCandidateScore<T extends DreamingShadowTrialCandidateInput> = {

51+

candidate: T;

52+

scoreBeforeShadowTrial: number;

53+

scoreAfterShadowTrial: number;

54+

shadowTrialScoreDelta: number;

55+

shadowTrialVerdict: DreamingShadowTrialVerdict;

56+

shadowTrialRecommendation: DreamingShadowTrialRecommendation;

57+

shadowTrialRiskFlags: string[];

58+

shadowTrialEvidenceRefs: string[];

59+

rejectedByShadowTrial: boolean;

60+

scoringAction: "report-only";

61+

};

62+3963

function normalizeRequiredText(value: string, label: string): string {

4064

const normalized = value.trim().replace(/\s+/g, " ");

4165

if (!normalized) {

@@ -49,6 +73,24 @@ function normalizeList(values: string[] | undefined, fallback: string): string[]

4973

return normalized.length > 0 ? normalized : [fallback];

5074

}

517576+

function normalizeDataList(values: string[] | undefined): string[] {

77+

return (values ?? []).map((value) => value.trim()).filter(Boolean);

78+

}

79+80+

function clampScore(value: number): number {

81+

if (!Number.isFinite(value)) {

82+

return 0;

83+

}

84+

return Math.max(0, Math.min(1, value));

85+

}

86+87+

function normalizeScoreDelta(value: number, fallback: number): number {

88+

if (!Number.isFinite(value)) {

89+

return fallback;

90+

}

91+

return Math.max(-1, Math.min(1, value));

92+

}

93+5294

export function resolveDreamingShadowTrialRecommendation(

5395

verdict: DreamingShadowTrialVerdict,

5496

): DreamingShadowTrialRecommendation {

@@ -88,6 +130,19 @@ function resolveReportContentHash(params: {

88130

return crypto.createHash("sha256").update(seed).digest("hex").slice(0, 12);

89131

}

90132133+

function resolveDreamingShadowTrialScoreDelta(

134+

verdict: DreamingShadowTrialVerdict,

135+

options?: DreamingShadowTrialScoreOptions,

136+

): number {

137+

if (verdict === "helpful") {

138+

return normalizeScoreDelta(options?.helpfulBoost ?? 0.04, 0.04);

139+

}

140+

if (verdict === "harmful") {

141+

return normalizeScoreDelta(options?.harmfulPenalty ?? -1, -1);

142+

}

143+

return normalizeScoreDelta(options?.neutralDelta ?? 0, 0);

144+

}

145+91146

export function defaultDreamingShadowTrialReportPath(params: {

92147

workspaceDir: string;

93148

candidate: string;

@@ -110,8 +165,8 @@ export function defaultDreamingShadowTrialReportPath(params: {

110165

candidateOutcome: normalizeRequiredText(params.candidateOutcome, "candidateOutcome"),

111166

verdict: params.verdict,

112167

reason: normalizeRequiredText(params.reason, "reason"),

113-

riskFlags: normalizeList(params.riskFlags, "none recorded"),

114-

evidenceRefs: normalizeList(params.evidenceRefs, "none supplied"),

168+

riskFlags: normalizeDataList(params.riskFlags),

169+

evidenceRefs: normalizeDataList(params.evidenceRefs),

115170

});

116171

return path.join(

117172

params.workspaceDir,

@@ -172,8 +227,8 @@ export function buildDreamingShadowTrialReport(

172227

const baselineOutcome = normalizeRequiredText(input.baselineOutcome, "baselineOutcome");

173228

const candidateOutcome = normalizeRequiredText(input.candidateOutcome, "candidateOutcome");

174229

const reason = normalizeRequiredText(input.reason, "reason");

175-

const riskFlags = normalizeList(input.riskFlags, "none recorded");

176-

const evidenceRefs = normalizeList(input.evidenceRefs, "none supplied");

230+

const riskFlags = normalizeDataList(input.riskFlags);

231+

const evidenceRefs = normalizeDataList(input.evidenceRefs);

177232

const recommendation = resolveDreamingShadowTrialRecommendation(input.verdict);

178233

const reportPath = resolveReportPath({

179234

workspaceDir: input.workspaceDir,

@@ -201,9 +256,9 @@ export function buildDreamingShadowTrialReport(

201256

`recommendation: ${recommendation}`,

202257

`reason: ${reason}`,

203258

"risk flags:",

204-

formatList(riskFlags),

259+

formatList(normalizeList(riskFlags, "none recorded")),

205260

"evidence refs:",

206-

formatList(evidenceRefs),

261+

formatList(normalizeList(evidenceRefs, "none supplied")),

207262

"promotion action: report-only",

208263

"",

209264

].join("\n");

@@ -224,6 +279,68 @@ export function buildDreamingShadowTrialReport(

224279

};

225280

}

226281282+

export function scoreDreamingShadowTrialCandidate<T extends DreamingShadowTrialCandidateInput>(

283+

candidate: T,

284+

report: Pick<

285+

DreamingShadowTrialReport,

286+

"verdict" | "recommendation" | "riskFlags" | "evidenceRefs"

287+

>,

288+

options?: DreamingShadowTrialScoreOptions,

289+

): DreamingShadowTrialCandidateScore<T> {

290+

const scoreBeforeShadowTrial = clampScore(candidate.score);

291+

const shadowTrialScoreDelta = resolveDreamingShadowTrialScoreDelta(report.verdict, options);

292+

const rejectedByShadowTrial = report.verdict === "harmful" || report.recommendation === "reject";

293+

const scoreAfterShadowTrial = rejectedByShadowTrial

294+

? 0

295+

: clampScore(scoreBeforeShadowTrial + shadowTrialScoreDelta);

296+297+

return {

298+

candidate,

299+

scoreBeforeShadowTrial,

300+

scoreAfterShadowTrial,

301+

shadowTrialScoreDelta,

302+

shadowTrialVerdict: report.verdict,

303+

shadowTrialRecommendation: report.recommendation,

304+

shadowTrialRiskFlags: normalizeDataList(report.riskFlags),

305+

shadowTrialEvidenceRefs: normalizeDataList(report.evidenceRefs),

306+

rejectedByShadowTrial,

307+

scoringAction: "report-only",

308+

};

309+

}

310+311+

export function rankDreamingShadowTrialCandidates<T extends DreamingShadowTrialCandidateInput>(

312+

candidates: readonly T[],

313+

reportsByCandidateKey: ReadonlyMap<string, DreamingShadowTrialReport>,

314+

options?: DreamingShadowTrialScoreOptions,

315+

): DreamingShadowTrialCandidateScore<T>[] {

316+

return candidates

317+

.map((candidate) => {

318+

const report = reportsByCandidateKey.get(candidate.key);

319+

if (!report) {

320+

const score = clampScore(candidate.score);

321+

return {

322+

candidate,

323+

scoreBeforeShadowTrial: score,

324+

scoreAfterShadowTrial: score,

325+

shadowTrialScoreDelta: 0,

326+

shadowTrialVerdict: "neutral" as const,

327+

shadowTrialRecommendation: "defer" as const,

328+

shadowTrialRiskFlags: ["not shadow-trialed"],

329+

shadowTrialEvidenceRefs: [],

330+

rejectedByShadowTrial: false,

331+

scoringAction: "report-only" as const,

332+

};

333+

}

334+

return scoreDreamingShadowTrialCandidate(candidate, report, options);

335+

})

336+

.toSorted((left, right) => {

337+

if (left.rejectedByShadowTrial !== right.rejectedByShadowTrial) {

338+

return left.rejectedByShadowTrial ? 1 : -1;

339+

}

340+

return right.scoreAfterShadowTrial - left.scoreAfterShadowTrial;

341+

});

342+

}

343+227344

export async function writeDreamingShadowTrialReport(

228345

input: DreamingShadowTrialInput & { workspaceDir: string },

229346

): Promise<DreamingShadowTrialReport> {