diff --git a/apps/ui/src/app/dashboard/[orgId]/[projectId]/activity/[logId]/log-detail-client.tsx b/apps/ui/src/app/dashboard/[orgId]/[projectId]/activity/[logId]/log-detail-client.tsx
index 5144981501..f676894f91 100644
--- a/apps/ui/src/app/dashboard/[orgId]/[projectId]/activity/[logId]/log-detail-client.tsx
+++ b/apps/ui/src/app/dashboard/[orgId]/[projectId]/activity/[logId]/log-detail-client.tsx
@@ -205,16 +205,19 @@ function formatDuration(ms: number) {
 
 // Selection reasons where the weighted-score formula is bypassed entirely, so
 // every provider's score is a hardcoded 0 placeholder rather than a real value.
+// "session-sticky" is intentionally excluded: it scores providers with the
+// normal weighted algorithm and pins the result for the session, so the logged
+// scores are real values worth surfacing. The all-zero fallback below hides
+// those scores when they couldn't be computed (no metrics available).
 const SCORE_BYPASSED_SELECTION_REASONS = new Set([
-	"session-sticky",
 	"random-exploration",
 	"price-only-no-metrics",
 ]);
 
 // The per-provider score only carries information when scoring actually ran.
-// Sticky/exploration/price-only paths emit 0 for every provider, and
-// "stable-preferred" can layer on top of a sticky pick (also all zeros), so
-// treat an all-zero set as "scoring did not run" regardless of the reason.
+// Exploration/price-only paths emit 0 for every provider, and "stable-preferred"
+// can layer on top of a sticky pick, so treat an all-zero set as "scoring did
+// not run" regardless of the reason.
 function isProviderScoreMeaningful(
 	selectionReason: string | null | undefined,
 	providerScores: { score: number }[],