@@ -3,6 +3,7 @@ import { HANDOFF_MODES, EXECUTION_HANDOFF_MODES, type ExecutionHandoffMode } fro
 import { normalizeGoalDraft } from "../../../../../loopx/control_plane/collaboration/goal_draft.js";
 import { parseTurnStep, type TurnStep } from "./turn-steps.js";
 import { z } from "zod";
+import {normalizeExploreMeasurements} from "../../../../../loopx/control_plane/capabilities/explore_measurements.js";
 import { actionSourceBasisSchema, todoUpdateSourceBasisSchema } from "./action-source-basis.js";
 
 import {
@@ -2665,6 +2666,7 @@ const exploreResultPageSchema = z.object({
     finding_id: z.string(), finding: z.string(), summary: z.string(), status: z.string(),
     node_id: z.string(), question: z.string(), scope: z.string(), agent_id: z.string(),
     evidence_refs: z.array(z.string()), last_updated_at: z.string(),
+    measurements: z.unknown().transform(normalizeExploreMeasurements).optional(),
     linked_todos: z.array(z.object({
       todo_id: z.string(), text: z.string(), status: z.string(), claimed_by: z.string(),
     })),
 
@@ -0,0 +1,38 @@
+import type {ExploreMeasurement, ExploreMeasurementFeedback} from "../../../../../../loopx/control_plane/capabilities/explore_measurements.js";
+
+export const formatExploreMeasurementValue = (value: number | null, zh: boolean) => value === null
+  ? (zh ? "未测量" : "Not measured")
+  : new Intl.NumberFormat(zh ? "zh-CN" : "en", {maximumSignificantDigits: 12}).format(value);
+
+/** Values and comparisons come from Explore's shared owner, never a UI scorer. */
+export function ExploreMeasurements({measurements, feedback = [], zh, title}: {
+  measurements: ExploreMeasurement[]; feedback?: ExploreMeasurementFeedback[]; zh: boolean; title?: string;
+}) {
+  if (!measurements.length) return null;
+  const number = (value: number | null) => formatExploreMeasurementValue(value, zh);
+  const label = title ?? (zh ? "量化结果" : "Measured results");
+  return <section className="explore-measurements" aria-label={label}>
+    <h4>{label}</h4>
+    {measurements.map((measurement, index) => {
+      const comparison = feedback.find(item => item.metric === measurement.metric && item.unit === measurement.unit
+        && item.higher_is_better === measurement.higher_is_better && item.comparison_scope === measurement.comparison_scope
+        && item.input_revision === measurement.input_revision && item.evaluation_ref === measurement.evaluation_ref);
+      const delta = comparison?.improvement;
+      return <div key={index}>
+        <p><strong>{measurement.metric} · {number(measurement.value)} {measurement.unit}</strong>
+          <small> · {measurement.higher_is_better ? (zh ? "越高越好" : "Higher is better") : (zh ? "越低越好" : "Lower is better")}</small></p>
+        {comparison ? <p>{zh ? "同口径 best" : "Best in this scope"}: {number(comparison.best?.value ?? null)}
+          {delta !== null && delta !== undefined ? <> · {zh ? "相对先前 best 的改善" : "Improvement over prior best"}: {delta > 0 ? "+" : ""}{number(delta)}</> : null}</p> : null}
+        <details><summary>{zh ? "评测范围与来源" : "Evaluation scope and sources"}</summary>
+          <p>{zh ? "比较范围" : "Comparison scope"}: {measurement.comparison_scope}<br/>
+            {zh ? "输入" : "Input"}: {measurement.input_revision}<br/>
+            {zh ? "评测" : "Evaluation"}: {measurement.evaluation_ref}<br/>
+            {zh ? "候选" : "Candidate"}: {measurement.candidate_ref}<br/>
+            {zh ? "结果引用" : "Result reference"}: {measurement.evidence_ref}</p>
+          {comparison?.best ? <p>{zh ? "best 候选 / 来源" : "Best candidate / source"}: {comparison.best.candidate_ref} · {comparison.best.evidence_ref}</p> : null}
+          <p>{zh ? "同一输入、评测和预算条件下的来源记录；best 不代表当前保留版本或验收通过。" : "Source-recorded values under the same input, evaluation and budget conditions; best does not establish the retained version or acceptance."}</p>
+        </details>
+      </div>;
+    })}
+  </section>;
+}
@@ -1,6 +1,7 @@
 import {useEffect, useRef, useState} from "react";
 import {RefreshCw} from "lucide-react";
 import {fetchExploreResults, type ExploreResultPage} from "../../data/chat";
+import {ExploreMeasurements} from "./explore-measurements";
 
 /** Read-only saved evidence, including history when Explore has been disabled. */
 export function GoalExploreResults({goalId, zh}: {goalId: string; zh: boolean}) {
@@ -44,6 +45,7 @@ export function GoalExploreResults({goalId, zh}: {goalId: string; zh: boolean})
       {selected ? <div ref={reader} tabIndex={-1} className="goal-team-result-reader" style={{overflowWrap: "anywhere"}}>
         <h4>{selected.question || selected.finding}</h4>
         {selected.scope ? <p>{zh ? "问题范围" : "Question scope"}: {selected.scope}</p> : null}
+        {selected.measurements ? <ExploreMeasurements measurements={selected.measurements} zh={zh}/> : null}
         <p style={{whiteSpace: "pre-wrap", overflowWrap: "anywhere"}}>{selected.summary}</p>
         <p>{zh ? "记录状态" : "Recorded status"}: {selected.status}</p>
         <p>{zh ? "记录者" : "Recorded by"}: {selected.agent_id || "—"}</p>
 
@@ -9,6 +9,7 @@ import {sourceReadApiUrl} from "../../data/local-status-query";
 import {ContextHelp} from "./context-help";
 import {groupUnlinkedHistoryQuestions, layoutExploreHistory, parseExploreHistoryPage, projectExploreHistory, type ExploreHistoryEvent} from "../../../../../../loopx/control_plane/presentation/explore_history.js";
 import "./goal-research-history.css";
+import {ExploreMeasurements, formatExploreMeasurementValue} from "./explore-measurements";
 
 type Selection = {nodes: string[]; day: string | null};
 type Lens = "timeline" | "graph";
@@ -78,7 +79,7 @@ export function GoalResearchHistory({goalId, active, zh, refreshRevision = 0, re
     const decision = sections?.decision;
     const execution = finding.execution_observation;
     const measured = (value: unknown) => value === null ? (zh ? "未测量" : "Not measured") : String(value);
-    return <>{sections ? <dl className="research-finding-sections">
+    return <>{finding.measurements ? <ExploreMeasurements measurements={finding.measurements} zh={zh}/> : null}{sections ? <dl className="research-finding-sections">
       <dt>{sectionCopy.observation}</dt><dd>{sections.observation}</dd>
       {decision ? <>
         <dt>{zh ? "决策" : "Decision"}</dt><dd><b className={`research-outcome is-${decision.outcome}`}>{outcomeCopy[decision.outcome] ?? decision.outcome}</b></dd>
@@ -99,7 +100,7 @@ export function GoalResearchHistory({goalId, active, zh, refreshRevision = 0, re
     </details> : null}</>;
   };
   const findingElement = (finding: ExploreHistoryEvent) => <div className="research-recorded-finding" key={finding.event_id}><strong>{finding.title}</strong><small>{timeLabel(finding.recorded_at)} · {zh ? "来源标记" : "Source label"} · {stateLabel(finding.status)}{finding.confidence !== null ? ` · ${zh ? "来源自报置信度" : "Source-reported confidence"} ${finding.confidence}` : ""}</small>{findingBody(finding)}{finding.evidence_refs.map(ref => <code key={ref}>{ref}</code>)}</div>;
-  const updateRecords = (events: readonly ExploreHistoryEvent[]) => <ol className="research-history-records">{events.map(event => <li key={event.event_id}><time>{timeLabel(event.recorded_at)}</time><span>{stateLabel(event.status)} · {event.title}{event.summary ? ` — ${event.summary}` : ""}</span>{event.parent_id ? <code>{copy.group}: {event.parent_id}</code> : null}{event.evidence_refs.map(ref => <code key={ref}>{ref}</code>)}{event.run_id ? <code>{event.run_id}</code> : null}</li>)}</ol>;
+  const updateRecords = (events: readonly ExploreHistoryEvent[]) => <ol className="research-history-records">{events.map(event => <li key={event.event_id}><time>{timeLabel(event.recorded_at)}</time><span>{stateLabel(event.status)} · {event.title}{event.summary ? ` — ${event.summary}` : ""}</span>{event.measurements ? <ExploreMeasurements measurements={event.measurements} zh={zh}/> : null}{event.parent_id ? <code>{copy.group}: {event.parent_id}</code> : null}{event.evidence_refs.map(ref => <code key={ref}>{ref}</code>)}{event.run_id ? <code>{event.run_id}</code> : null}</li>)}</ol>;
 
   useEffect(() => {
     const pop = () => setSelection(selectedFromUrl(goalId));
@@ -147,6 +148,12 @@ export function GoalResearchHistory({goalId, active, zh, refreshRevision = 0, re
     catch (failure) {return {value: null, error: String(failure)};}
   }, [read.events, goalId]);
   const value = history.value;
+  const measurementFeedback = value?.measurementFeedback ?? [];
+  const measurementSummary = (nodeId: string) => {
+    const feedback = measurementFeedback.filter(item => item.node_id === nodeId);
+    return feedback.length ? <ExploreMeasurements title={zh ? "最近记录与历史最佳" : "Latest recorded and historical best"}
+      measurements={feedback.map(item => item.latest.measurement)} feedback={feedback} zh={zh}/> : null;
+  };
   const grouped = useMemo(() => groupUnlinkedHistoryQuestions(value?.nodes ?? [], value?.edges ?? []), [value]);
   const layout = useMemo(() => layoutExploreHistory(grouped.nodes, value?.edges ?? []), [grouped, value]);
   const timeline = useMemo(() => value ? buildExploreTimeline(value.nodes, value.edges, value.events) : null, [value]);
@@ -223,11 +230,15 @@ export function GoalResearchHistory({goalId, active, zh, refreshRevision = 0, re
               return <path key={`parent:${item.node.id}`} className="is-grouping" d={`M${parent.x + 206},${parent.y + 47} C${parent.x + 226},${parent.y + 47} ${item.x - 20},${item.y + 47} ${item.x},${item.y + 47}`} />;
             })}
           </svg>
-          {layout.map(({node, x, y}) => <button key={node.id} type="button" className={`research-history-node is-${node.latest.status}${selection.nodes.includes(node.id) ? " is-selected" : ""}`}
+          {layout.map(({node, x, y}) => {
+            const measured = measurementFeedback.find(item => item.node_id === node.id);
+            return <button key={node.id} type="button" className={`research-history-node is-${node.latest.status}${selection.nodes.includes(node.id) ? " is-selected" : ""}`}
             style={{left: x, top: y}} aria-pressed={selection.nodes.includes(node.id)} aria-label={`${node.latest.title} · ${stateLabel(node.latest.status)}`} onClick={() => choose(node.id)}>
             <span className="research-node-meta">{node.latest.node_kind === "experiment" ? <FlaskConical size={13} aria-hidden/> : <Circle size={12} aria-hidden/>}{kindLabel(node.latest.node_kind)}<small>{stateLabel(node.latest.status)}</small></span>
-            <strong>{node.latest.title}</strong><span className="research-node-count"><time>{timeLabel(node.updates[0]?.recorded_at)}</time>{node.updates.length > 1 ? ` · ${node.updates.length} ${copy.changes}` : ""}{node.findings.length ? ` · ${node.findings.length} ${copy.findings}` : ""}</span>
-          </button>)}
+            <strong>{node.latest.title}</strong>{measured ? <span className="research-node-count" title={`${measured.metric} (${measured.unit}) · ${measured.comparison_scope} · ${measured.evaluation_ref}`}>
+              {measured.metric} {formatExploreMeasurementValue(measured.latest.measurement.value, zh)} · best {formatExploreMeasurementValue(measured.best?.value ?? null, zh)} {measured.higher_is_better ? "↑" : "↓"}
+            </span> : <span className="research-node-count"><time>{timeLabel(node.updates[0]?.recorded_at)}</time>{node.updates.length > 1 ? ` · ${node.updates.length} ${copy.changes}` : ""}{node.findings.length ? ` · ${node.findings.length} ${copy.findings}` : ""}</span>}
+          </button>})}
           {grouped.groups.map((group, index) => <button type="button" key={`day:${group.day}`} className={`research-history-node research-history-group${selectedGroup === group ? " is-selected" : ""}`} style={{left: (index % 4) * 230 + 24, top: clusterY + Math.floor(index / 4) * 124}}
             aria-pressed={selectedGroup === group} onClick={() => select({nodes: [], day: group.day})}>
             <span className="research-node-meta"><Circle size={12} aria-hidden/>{zh ? "未关联问题" : "Unlinked questions"}</span>
@@ -239,6 +250,7 @@ export function GoalResearchHistory({goalId, active, zh, refreshRevision = 0, re
         <p className="research-detail-kind"><GitBranch size={14} aria-hidden/>{kindLabel(selected.latest.node_kind)} · {stateLabel(selected.latest.status)}</p>
         <h3>{selected.latest.title}</h3><p>{selected.latest.summary || (zh ? "此节点尚无详细说明。" : "No detail was recorded for this node.")}</p>
         {selected.latest.blocked_reason ? <p className="research-history-error">{selected.latest.blocked_reason}</p> : null}
+        {measurementSummary(selected.id)}
         <h4>{copy.evidence}</h4>
         {selected.findings.length ? selected.findings.map(findingElement) : <p>{zh ? "此节点没有附属结论记录。" : "No finding is attached to this node."}</p>}
         <h4>{copy.versions}</h4>{selected.latest.evidence_refs.length ? selected.latest.evidence_refs.map(ref => <code key={ref}>{ref}</code>) : <p>{zh ? "产物版本与来源引用未记录。" : "Artifact version and source references were not recorded."}</p>}
@@ -249,6 +261,7 @@ export function GoalResearchHistory({goalId, active, zh, refreshRevision = 0, re
       {selectedMembers?.length ? <div className="research-history-detail" ref={detail} role="region" aria-label={selectedGroup ? (zh ? "按日期收纳的未关联问题" : "Unlinked questions by date") : (zh ? "相邻研究项的全部记录" : "All records for nearby research items")} tabIndex={-1} onKeyDown={event => {if (event.key === "Escape") {event.stopPropagation(); choose("", false); graph.current?.focus();}}}>
         <h3>{selectedGroup ? `${selectedGroup.day || copy.unknown} · ${zh ? "未关联问题" : "Unlinked questions"}` : (zh ? `相邻研究项 · ${selectedMembers.length}` : `Nearby research items · ${selectedMembers.length}`)}</h3>
         {selectedMembers.map(node => <article className="research-history-group-member" key={node.id} data-research-node={node.id}><h4>{node.latest.title}</h4><p>{stateLabel(node.latest.status)} · {node.latest.summary}</p>{node.latest.evidence_refs.map(ref => <code key={ref}>{ref}</code>)}
+          {measurementSummary(node.id)}
           {node.findings.map(findingElement)}
           {updateRecords(node.revisions)}
         </article>)}
 
@@ -12,6 +12,46 @@ This records producer evidence, not verified execution or utility. Findings,
 replan context and the existing research-history detail retain it; omission
 preserves old events. See the [local probe producer](../../../benchmark/runtime/EXECUTION_OBSERVATIONS.md).
 
+## Quantitative findings
+
+An ordinary result attachment can include `measurements`, saved on its existing
+node-linked finding. The graph and research-history detail retain the values;
+`explore turn-context` supplies bounded `measurement_feedback` to the next turn.
+Manual capture uses `explore finding --node <id> --evidence-ref result:a
+--measurements-json measurements.json`. The file contains one to eight items:
+
+```json
+[{"metric":"quality","value":8,"unit":"points","higher_is_better":true,
+  "comparison_scope":"validation:fixed-budget-v1","input_revision":"dataset-v1",
+  "evaluation_ref":"evaluator:v1","candidate_ref":"candidate:a","evidence_ref":"result:a"}]
+```
+
+Each result reference must belong to the finding's `evidence_refs`; ordinary
+attachments also require the same `input_revision`. Use a new finding id per
+evaluated candidate, and reuse an id only to correct that finding. Missing values
+are explicit `null`; zero and negative values remain valid measurements.
+
+Comparisons require the same node, metric, unit, direction, comparison scope,
+input revision and evaluator. Producers must change the scope when task,
+model/resource budget or evaluation conditions change, and separate native
+evaluation from local proxies. Explore validates the declared comparison keys,
+not their truth. It computes the latest recorded measurement, historical best
+candidate/source and direction-adjusted improvement against the prior best,
+before applying the context's display limit. Latest follows source-log append
+order, including tied or backfilled timestamps; its ordinal is derived, not
+another persisted fact. Corrected findings replace earlier
+versions in comparisons; original versions remain in history. Counts are records,
+not independent trials. Feedback pages reuse `--result-limit` (1..20, default 3),
+`--result-offset`, `--result-revision` and the optional `--result-node` filter.
+Follow `measurement_feedback_page.next_command` for additional groups, including
+multiple metrics or conditions on one node. The shared revision binds all compared
+sources, including manual findings; a changed source requires restarting the read.
+The next-turn guidance uses this feedback to investigate
+regressions or revisit better candidates before repeating work. Best does not
+identify the current retained candidate, establish causality, satisfy acceptance
+or authorize execution. Actual adoption and outcome benefit require subsequent
+work and evidence; automatic branch rankings remain unchanged.
+
 ## At a Glance
 
 Explore Harness is a supported, default-off optional capability for
 
@@ -9,6 +9,38 @@
 不认证执行真实性或收益；finding、replan 和既有研究历程详情保留相同信息，缺省仍兼容
 旧事件。参见[本地探针生产路径](../../../benchmark/runtime/EXECUTION_OBSERVATIONS.md)。
 
+## 量化结论
+
+普通结果附件可附带 `measurements`，保存到既有节点所属的 finding。关系图和
+研究历程详情保留分数；`explore turn-context` 将有界的 `measurement_feedback`
+交给下一轮。手动记录使用 `explore finding --node <id> --evidence-ref result:a
+--measurements-json measurements.json`，文件包含一至八项：
+
+```json
+[{"metric":"quality","value":8,"unit":"points","higher_is_better":true,
+  "comparison_scope":"validation:fixed-budget-v1","input_revision":"dataset-v1",
+  "evaluation_ref":"evaluator:v1","candidate_ref":"candidate:a","evidence_ref":"result:a"}]
+```
+
+结果引用必须属于 finding 的 `evidence_refs`；普通附件还要求相同 `input_revision`。
+每个受评测候选使用新 finding id，仅纠正同一结论时复用 id。未测量显式记为 `null`，
+零分和负分仍为有效测量。
+
+只有节点、指标、单位、方向、比较范围、输入版本和评测器全部一致时才比较。
+任务、模型或资源预算、评测条件改变时，生产者必须更换比较范围；原生评测与本地
+代理指标分别记录。Explore 校验声明的比较键，不认证声明真实性。它在上下文裁剪前
+计算最近记录的测量、历史最佳候选及来源、相对先前最佳的方向归一化改善幅度。
+最近记录按源日志追加顺序确定；时间相同或事后回填不改变此顺序，序号仅从日志
+派生，不另行持久化。
+纠正结论会在比较中替换旧版本，原始版本仍可在历史中读取；计数表示记录数，
+不表示独立试验次数。反馈分页复用 `--result-limit`（1..20，默认 3）、
+`--result-offset`、`--result-revision` 和可选的 `--result-node` 过滤。
+沿 `measurement_feedback_page.next_command` 读取其余组，同一节点的多个指标或条件
+也可完整翻页。共同 revision 绑定全部比较来源，包含手动 finding；来源变化须从头读取。
+下一轮指导使用这些反馈排查退步、回看更好候选，再决定是否
+重复尝试。best 不代表当前保留版本、因果收益、验收通过或执行授权；是否实际采用
+并产生收益，仍需后续工作和证据确认。自动分支排序保持原有行为。
+
 ## 概览
 
 LoopX Explore 是面向长程探索目标（软件研究、安全攻击面测绘、领域调研）的受支持、默认关闭的可选能力。它把“到处看看”变成有界、可观测、有门控的流程，包含三根支柱：