diff --git a/app/(chat)/api/chat/route.ts b/app/(chat)/api/chat/route.ts
index cc590ef4e..380443b7e 100644
--- a/app/(chat)/api/chat/route.ts
+++ b/app/(chat)/api/chat/route.ts
@@ -3002,7 +3002,7 @@ export async function POST(request: Request) {
           /用户已选中\s*\d+\s*张图片/.test(autoContextText);
         const hasExplicitTargetHint =
           requestTargetResolution.resolutionMode === "text_reference" ||
-          /第[一二三四五六七八九十\d]+张|左|右|上|下|中间|选中/.test(requestTextContent);
+          /第[一二三四五六七八九十\d]+张|左边|左侧|左上|左下|右边|右侧|右上|右下|上面|上方|下面|下方|中间|居中|中央|选中/.test(requestTextContent);
         const hasActionableEditIntent = /改|调整|优化|换|替换|去掉|抠|重绘|增强|修|处理|按照|按|保留|统一|色差|背景|清晰|风格/.test(
           requestTextContent
         );
diff --git a/lib/ai/context-memory/target-resolution.ts b/lib/ai/context-memory/target-resolution.ts
index 1ee5bc629..f27745929 100644
--- a/lib/ai/context-memory/target-resolution.ts
+++ b/lib/ai/context-memory/target-resolution.ts
@@ -103,11 +103,11 @@ function pickBySpatialHint(text: string, refs: ResolvedImageRef[]): ResolvedImag
   if (!refs.length) return [];
 
   const normalizedText = text.replace(/\s+/g, "");
-  const hasLeft = normalizedText.includes("左");
-  const hasRight = normalizedText.includes("右");
-  const hasTop = normalizedText.includes("上");
-  const hasBottom = normalizedText.includes("下");
-  const hasCenter = normalizedText.includes("中间") || normalizedText.includes("中");
+  const hasLeft = /左边|左侧|左面|左上|左下|靠左/.test(normalizedText);
+  const hasRight = /右边|右侧|右面|右上|右下|靠右/.test(normalizedText);
+  const hasTop = /上面|上方|上边|上侧|顶部|靠上|最上/.test(normalizedText);
+  const hasBottom = /下面|下方|下边|下侧|底部|靠下|最下/.test(normalizedText);
+  const hasCenter = /中间|居中|正中|中央/.test(normalizedText);
 
   if (!(hasLeft || hasRight || hasTop || hasBottom || hasCenter)) return [];