diff --git a/backend/app/main.py b/backend/app/main.py
index 2dbcfa1..4f0e642 100644
--- a/backend/app/main.py
+++ b/backend/app/main.py
@@ -280,7 +280,7 @@ def runs():
             improvement
             for response in responses
             for improvement in response.get("improvements", [])
-            if isinstance(improvement, dict)
+            if isinstance(improvement, dict) and not str(improvement.get("known_issue_id") or "").strip()
         ]
         issues = [
             issue
diff --git a/backend/app/store.py b/backend/app/store.py
index c74e7dc..a344171 100644
--- a/backend/app/store.py
+++ b/backend/app/store.py
@@ -2429,7 +2429,7 @@ def _known_unresolved_issues(self, run: Run, *, limit: int = 12) -> list[dict[st
             candidates.append(run)
         values: list[dict[str, str]] = []
         for source_run in candidates:
-            for result in source_run.supervisor_results:
+            for record_index, result in enumerate(source_run.supervisor_results):
                 if (
                     not isinstance(result, dict)
                     or result.get("stage", "issue_assessment") != "issue_assessment"
@@ -2446,9 +2446,11 @@ def _known_unresolved_issues(self, run: Run, *, limit: int = 12) -> list[dict[st
                 for index, proposal in enumerate(proposals):
                     if not isinstance(proposal, dict):
                         continue
-                    proposal_id = f"{source_run.id}:{result.get('iteration', 0)}:{index}"
+                    proposal_id = f"{source_run.id}:{result.get('iteration', 0)}:{record_index}:{index}"
                     if proposals is reported:
-                        proposal_id = f"{source_run.id}:{result.get('iteration', 0)}:issue:{index}"
+                        proposal_id = (
+                            f"{source_run.id}:{result.get('iteration', 0)}:{record_index}:issue:{index}"
+                        )
                     management = records.get(proposal_id, {})
                     if management.get("status") in {"resolved", "deleted"}:
                         continue
@@ -3395,6 +3397,11 @@ def parse_timestamp(value: object) -> datetime | None:
                 improvements = (
                     response.get("improvements", []) if isinstance(response.get("improvements"), list) else []
                 )
+                improvements = [
+                    item
+                    for item in improvements
+                    if isinstance(item, dict) and not str(item.get("known_issue_id") or "").strip()
+                ]
                 issues = (
                     response.get("reported_issues", [])
                     if isinstance(response.get("reported_issues"), list)
@@ -3575,7 +3582,7 @@ def active_evaluations(self, runs: list[Run] | None = None) -> list[dict[str, An
                 improvement
                 for response in responses
                 for improvement in response.get("improvements", [])
-                if isinstance(improvement, dict)
+                if isinstance(improvement, dict) and not str(improvement.get("known_issue_id") or "").strip()
             ]
             issues = [
                 issue
@@ -4608,7 +4615,14 @@ def supervisor_result(result: object) -> object:
                 reported_at = now().isoformat()
                 for improvement in result["improvements"]:
                     improvement.setdefault("reported_at", reported_at)
-                    improvement.setdefault("effect_score", (result.get("evaluation") or {}).get("score"))
+                    item_evaluation = improvement.get("evaluation")
+                    item_score = item_evaluation.get("score") if isinstance(item_evaluation, dict) else None
+                    if improvement.get("effect_score") is None:
+                        improvement["effect_score"] = (
+                            item_score
+                            if item_score is not None
+                            else (result.get("evaluation") or {}).get("score")
+                        )
                     improvement.setdefault(
                         "attempted", improvement.get("status") in {"adopted", "accepted", "rejected"}
                     )
diff --git a/backend/tests/test_api.py b/backend/tests/test_api.py
index 3055c00..5580605 100644
--- a/backend/tests/test_api.py
+++ b/backend/tests/test_api.py
@@ -1025,7 +1025,13 @@ def test_active_evaluations_count_feedback_across_all_iterations(monkeypatch):
             {
                 "iteration": 1,
                 "response": {
-                    "improvements": [{"title": "Add refund intake", "status": "accepted"}],
+                    "improvements": [
+                        {"title": "Add refund intake", "status": "accepted"},
+                        {
+                            "known_issue_id": "feedback-history-run:0:0:0",
+                            "evidence": "The previously reported issue remains visible.",
+                        },
+                    ],
                     "reported_issues": [{"title": "Missing refund details"}],
                 },
             },
@@ -1738,7 +1744,7 @@ def complete(self, _settings, prompt):
                 {
                     "improvements": [
                         {
-                            "known_issue_id": "known-issue-run:1:0",
+                            "known_issue_id": "known-issue-run:1:0:0",
                             "evidence": "The basis is still absent in this iteration.",
                             "evaluation": {"score": 7, "approval": "pending", "summary": "Known issue."},
                         }
@@ -1769,10 +1775,12 @@ def complete(self, _settings, prompt):
     store._complete_supervision(run.id)
 
     assert "# Known unresolved issues" in captured_prompts[0]
-    assert "known-issue-run:1:0" in captured_prompts[0]
+    assert "known-issue-run:1:0:0" in captured_prompts[0]
     assert [item["title"] for item in store.proposal_lifecycles()] == [
         "Portfolio calculation basis is unclear"
     ]
+    retained_evidence = store._load(run.id).supervisor_results[-1]["response"]["improvements"][0]
+    assert retained_evidence["effect_score"] == 7
 
 
 def test_runner_context_uses_the_supplied_model_profile_without_exposing_its_secret(tmp_path, monkeypatch):
@@ -2668,6 +2676,12 @@ def test_issue_management_separates_multiple_supervisor_records_in_one_iteration
     by_title = {item["title"]: item for item in store.issue_management_items()}
     assert by_title["First issue"]["comments"][0]["body"] == "Only the first issue"
     assert by_title["Second issue"]["comments"] == []
+    assert {issue["id"] for issue in store._known_unresolved_issues(store.runs()[0])} == {
+        "run-duplicate-assessments:1:0:0",
+        "run-duplicate-assessments:1:1:0",
+        "run-duplicate-assessments:2:2:issue:0",
+        "run-duplicate-assessments:2:3:issue:0",
+    }
 
 
 def test_hello_accepts_unsaved_profile_settings():
diff --git a/frontend/src/features/evaluations/page.tsx b/frontend/src/features/evaluations/page.tsx
index d525b12..1a48b58 100644
--- a/frontend/src/features/evaluations/page.tsx
+++ b/frontend/src/features/evaluations/page.tsx
@@ -658,6 +658,7 @@ export function EvaluationsPage({
       issueStatuses = new Set<string>();
     for (const record of selected?.supervisor_results ?? []) {
       for (const improvement of record.response?.improvements ?? []) {
+        if (String(improvement.known_issue_id ?? "").trim()) continue;
         if (improvement.status) improvementStatuses.add(String(improvement.status));
       }
       for (const issue of record.response?.reported_issues ?? []) {
@@ -681,7 +682,9 @@ export function EvaluationsPage({
         recordEvaluation = response?.evaluation,
         decision = recordEvaluation?.approval ?? "no_response",
         score = recordEvaluation?.score,
-        improvements = response?.improvements ?? [],
+        improvements = (response?.improvements ?? []).filter(
+          (improvement) => !String(improvement.known_issue_id ?? "").trim(),
+        ),
         issues = response?.reported_issues ?? [];
       const inRange =
         (!resultIterationFrom || record.iteration >= Number(resultIterationFrom)) &&
@@ -760,7 +763,9 @@ export function EvaluationsPage({
     resultTranslationIds.every((templateId) => Boolean(resultTranslations.content(templateId)));
   const resultTranslationsLoading = resultTranslations.isLoading(resultTranslationIds);
   const resultImprovements = resultRecords.flatMap((record) =>
-      (translateResultResponse(record)?.improvements ?? []).map((item) => ({
+      (translateResultResponse(record)?.improvements ?? [])
+        .filter((item) => !String(item.known_issue_id ?? "").trim())
+        .map((item) => ({
         ...item,
         __iteration: record.iteration,
       })),
diff --git a/frontend/src/features/evaluations/run-detail-result-panel.tsx b/frontend/src/features/evaluations/run-detail-result-panel.tsx
index 0cb99d5..029f604 100644
--- a/frontend/src/features/evaluations/run-detail-result-panel.tsx
+++ b/frontend/src/features/evaluations/run-detail-result-panel.tsx
@@ -28,8 +28,11 @@ function browserJourney(result?: Record<string, unknown>) {
 
 function ResultList({ items, kind, locale, empty, l }: { items: RecordItem[]; kind: "improvement" | "issue"; locale: Locale; empty: string; l: Messages }) {
   return <div className="result-items result-items--scrollable">{items.length ? items.map((item, index) => {
-    const status = String(item.status ?? "—"), severity = String(item.severity ?? "—"), reportedAt = typeof item.reported_at === "string" ? item.reported_at : undefined;
-    return <article className="result-row" key={index}><time className="result-row__time">{time(locale, reportedAt)}</time><div className="result-row__body"><strong>{String(item.title ?? "—")}</strong><p>{String(kind === "improvement" ? (item.rationale ?? "—") : (item.evidence ?? "—"))}</p></div><div className="result-row__metrics"><span><small>{l.resultIteration}</small><b>#{String(item.__iteration ?? "—")}</b></span>{kind === "improvement" ? <><span><small>{l.resultStatus}</small><b className={`decision decision--${status}`}>{status}</b></span><span><small>{l.resultScore}</small><b>{String(item.effect_score ?? item.score ?? "—")}</b></span><span><small>{l.resultAttempted}</small><b>{item.attempted === true || ["adopted", "accepted"].includes(status) ? l.resultYes : l.resultNo}</b></span></> : <><span><small>{l.resultSeverity}</small><b className={`decision decision--${severity}`}>{severity}</b></span><span><small>{l.resultStatus}</small><b className={`decision decision--${status}`}>{status}</b></span></>}</div></article>;
+    const status = String(item.status ?? "—"), severity = String(item.severity ?? "—"), reportedAt = typeof item.reported_at === "string" ? item.reported_at : undefined,
+      evaluation = item.evaluation,
+      evaluationScore = evaluation && typeof evaluation === "object" ? (evaluation as RecordItem).score : undefined,
+      score = item.effect_score ?? item.score ?? evaluationScore ?? "—";
+    return <article className="result-row" key={index}><time className="result-row__time">{time(locale, reportedAt)}</time><div className="result-row__body"><strong>{String(item.title ?? "—")}</strong><p>{String(kind === "improvement" ? (item.rationale ?? "—") : (item.evidence ?? "—"))}</p></div><div className="result-row__metrics"><span><small>{l.resultIteration}</small><b>#{String(item.__iteration ?? "—")}</b></span>{kind === "improvement" ? <><span><small>{l.resultStatus}</small><b className={`decision decision--${status}`}>{status}</b></span><span><small>{l.resultScore}</small><b>{String(score)}</b></span><span><small>{l.resultAttempted}</small><b>{item.attempted === true || ["adopted", "accepted"].includes(status) ? l.resultYes : l.resultNo}</b></span></> : <><span><small>{l.resultSeverity}</small><b className={`decision decision--${severity}`}>{severity}</b></span><span><small>{l.resultStatus}</small><b className={`decision decision--${status}`}>{status}</b></span></>}</div></article>;
   }) : <p className="hint result-empty">{empty}</p>}</div>;
 }