Update app.py

app.py

@@ -43,11 +43,16 @@ llm = ChatOpenAI(model="gpt-4o", temperature=0)
 #llm_with_tools = llm.bind_tools(tools, parallel_tool_calls=False)
 
 # System message
-sys_msg = SystemMessage(content="You are an AI assistant completing benchmark questions. Assume fictional inputs (e.g., video, image, audio, code) and do your best to guess the answer accurately. Always answer concisely and confidently.")
+sys_msg = SystemMessage(
+    content="You are a precise and silent agent. Return only the final answer in lowercase. Never include explanations or restate the question. For example: just return 'right', '5', or 'extremely'."
+)
    
 def assistant(state: MessagesState):
     return {"messages": [llm.invoke([sys_msg] + state["messages"])]}
 
+def clean_output(text: str) -> str:
+    return text.strip().lower().strip('"').strip(".")
+
 
 # Build graph
 builder = StateGraph(MessagesState)
@@ -74,7 +79,9 @@ class BasicAgent:
         print(f"Agent received question (first 50 chars): {question[:50]}...")
         messages = [HumanMessage(content=question)]
         result = self.graph.invoke({"messages": messages})
-        answer = result['messages'][-1].content.strip().lower().strip('"').rstrip('.')
+        raw = result['messages'][-1].content
+        answer = clean_output(raw)
+
         return answer 
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):