diff --git a/backend/app/prompts/expert_transformation_prompt.py b/backend/app/prompts/expert_transformation_prompt.py
index 6cc304f..60d136e 100644
--- a/backend/app/prompts/expert_transformation_prompt.py
+++ b/backend/app/prompts/expert_transformation_prompt.py
@@ -10,20 +10,40 @@ def get_expert_generation_prompt(
     custom_experts: Optional[List[str]] = None
 ) -> str:
     """Step 0: 生成專家團隊（不依賴主題，純隨機多元）"""
+    import time
+    import random
+
     custom_text = ""
     if custom_experts and len(custom_experts) > 0:
         custom_text = f"（已指定：{', '.join(custom_experts[:expert_count])}）"
 
+    # 加入時間戳和隨機數來增加多樣性
+    seed = int(time.time() * 1000) % 10000
+    diversity_hints = [
+        "冷門、非主流、跨領域",
+        "罕見職業、新興領域、邊緣學科",
+        "非傳統、創新、小眾專業",
+        "未來趨向、實驗性、非常規",
+        "跨文化、混合領域、獨特視角"
+    ]
+    hint = random.choice(diversity_hints)
+
     return f"""/no_think
 隨機組建 {expert_count} 個來自完全不同領域的專家團隊{custom_text}。
 
+【創新要求】（隨機種子:{seed}）
+- 優先選擇{hint}的專家
+- 避免常見職業（如醫生、工程師、教師、律師等）
+- 每個專家必須來自完全不相關的領域
+- 越罕見、越創新越好
+
 回傳 JSON：
 {{"experts": [{{"id": "expert-0", "name": "職業", "domain": "領域", "perspective": "角度"}}, ...]}}
 
 規則：
 - id 為 expert-0 到 expert-{expert_count - 1}
 - name 填寫職業名稱（非人名），2-5字
-- 各專家的 domain 必須來自截然不同的領域，越多元越好"""
+- domain 要具體且獨特，不可重複類型"""
 
 
 def get_expert_keyword_generation_prompt(
@@ -48,31 +68,20 @@ def get_expert_keyword_generation_prompt(
 共需 {len(experts) * keywords_per_expert} 個關鍵字。"""
 
 
-def get_expert_batch_description_prompt(
+def get_single_description_prompt(
     query: str,
-    category: str,
-    expert_keywords: List[dict]  # List[ExpertKeyword]
+    keyword: str,
+    expert_id: str,
+    expert_name: str,
+    expert_domain: str
 ) -> str:
-    """Step 2: 批次生成專家關鍵字的描述"""
-    keywords_info = ", ".join([
-        f"{kw['expert_name']}:{kw['keyword']}"
-        for kw in expert_keywords
-    ])
-
-    # 建立 keyword -> (expert_id, expert_name) 的對照
-    keyword_expert_map = ", ".join([
-        f"{kw['keyword']}→{kw['expert_id']}/{kw['expert_name']}"
-        for kw in expert_keywords
-    ])
-
+    """Step 2: 為單一關鍵字生成描述"""
     return f"""/no_think
 物件：「{query}」
-關鍵字（專家:詞彙）：{keywords_info}
-對照：{keyword_expert_map}
+專家：{expert_name}（{expert_domain}）
+關鍵字：{keyword}
 
-為每個關鍵字生成創新描述（15-30字），說明如何將該概念應用到「{query}」上。
+從這位專家的視角，生成一段創新應用描述（15-30字），說明如何將「{keyword}」的概念應用到「{query}」上。
 
 回傳 JSON：
-{{"descriptions": [{{"keyword": "詞彙", "expert_id": "expert-X", "expert_name": "名稱", "description": "應用描述"}}, ...]}}
-
-共需 {len(expert_keywords)} 個描述。"""
+{{"description": "應用描述"}}"""
diff --git a/backend/app/routers/expert_transformation.py b/backend/app/routers/expert_transformation.py
index 8a0d737..a88d3ad 100644
--- a/backend/app/routers/expert_transformation.py
+++ b/backend/app/routers/expert_transformation.py
@@ -17,7 +17,7 @@ from ..models.schemas import (
 from ..prompts.expert_transformation_prompt import (
     get_expert_generation_prompt,
     get_expert_keyword_generation_prompt,
-    get_expert_batch_description_prompt,
+    get_single_description_prompt,
 )
 from ..services.llm_service import ollama_provider, extract_json_from_response
 
@@ -119,34 +119,48 @@ async def generate_expert_transformation_events(
             yield f"event: error\ndata: {json.dumps({'error': '無法生成關鍵字'}, ensure_ascii=False)}\n\n"
             return
 
-        # ========== Step 2: Generate descriptions for each expert keyword ==========
-        yield f"event: description_start\ndata: {json.dumps({'message': '為專家關鍵字生成創新應用描述...'}, ensure_ascii=False)}\n\n"
+        # ========== Step 2: Generate descriptions one by one ==========
+        yield f"event: description_start\ndata: {json.dumps({'message': '為專家關鍵字生成創新應用描述...', 'total': len(all_expert_keywords)}, ensure_ascii=False)}\n\n"
 
         descriptions: List[ExpertTransformationDescription] = []
 
-        try:
-            desc_prompt = get_expert_batch_description_prompt(
-                query=request.query,
-                category=request.category,
-                expert_keywords=[kw.model_dump() for kw in all_expert_keywords]
-            )
-            logger.info(f"Description prompt: {desc_prompt[:300]}")
+        # Build expert lookup for domain info
+        expert_lookup = {exp.id: exp for exp in experts}
 
-            desc_response = await ollama_provider.generate(
-                desc_prompt, model=model, temperature=temperature
-            )
-            logger.info(f"Description response: {desc_response[:500]}")
+        for idx, kw in enumerate(all_expert_keywords):
+            try:
+                expert = expert_lookup.get(kw.expert_id)
+                expert_domain = expert.domain if expert else ""
 
-            desc_data = extract_json_from_response(desc_response)
-            descriptions_raw = desc_data.get("descriptions", [])
+                desc_prompt = get_single_description_prompt(
+                    query=request.query,
+                    keyword=kw.keyword,
+                    expert_id=kw.expert_id,
+                    expert_name=kw.expert_name,
+                    expert_domain=expert_domain
+                )
 
-            for desc in descriptions_raw:
-                if isinstance(desc, dict) and all(k in desc for k in ["keyword", "expert_id", "expert_name", "description"]):
-                    descriptions.append(ExpertTransformationDescription(**desc))
+                desc_response = await ollama_provider.generate(
+                    desc_prompt, model=model, temperature=temperature
+                )
 
-        except Exception as e:
-            logger.warning(f"Failed to generate descriptions: {e}")
-            # Continue without descriptions - at least we have keywords
+                desc_data = extract_json_from_response(desc_response)
+                desc_text = desc_data.get("description", "")
+
+                if desc_text:
+                    descriptions.append(ExpertTransformationDescription(
+                        keyword=kw.keyword,
+                        expert_id=kw.expert_id,
+                        expert_name=kw.expert_name,
+                        description=desc_text
+                    ))
+
+                # Send progress update
+                yield f"event: description_progress\ndata: {json.dumps({'current': idx + 1, 'total': len(all_expert_keywords), 'keyword': kw.keyword}, ensure_ascii=False)}\n\n"
+
+            except Exception as e:
+                logger.warning(f"Failed to generate description for '{kw.keyword}': {e}")
+                # Continue with next keyword
 
         yield f"event: description_complete\ndata: {json.dumps({'count': len(descriptions)}, ensure_ascii=False)}\n\n"
 
diff --git a/frontend/src/App.tsx b/frontend/src/App.tsx
index 9c28d1f..9282250 100644
--- a/frontend/src/App.tsx
+++ b/frontend/src/App.tsx
@@ -33,7 +33,7 @@ function App() {
 
   // Transformation Agent settings
   const [transformModel, setTransformModel] = useState<string>('');
-  const [transformTemperature, setTransformTemperature] = useState<number>(0.7);
+  const [transformTemperature, setTransformTemperature] = useState<number>(0.95);
   const [expertConfig, setExpertConfig] = useState<{
     expert_count: number;
     keywords_per_expert: number;