Improve answer quality: cleanup + stricter format prompt

- _clean_answer(): strip quotes, trailing punctuation, $ signs, placeholders,
normalize list spacing ("a, b" -> "a,b")
- Prompt: no IOC codes, no quotes, no placeholder [answer], no spaces in lists
- Prompt: try GAIA benchmark search for unanswerable questions

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app.py

@@ -13,23 +13,25 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 SYSTEM_PROMPT = """You are a general AI assistant. Answer GAIA benchmark questions accurately.
 
 Tool usage:
-- duckduckgo_search: look up facts, names, numbers, events. Use specific queries. Try multiple searches.
-- Python_REPL: calculations, data analysis, string manipulation. ALWAYS use print() to output results.
+- duckduckgo_search: look up facts, names, numbers, events. Use specific queries. Try multiple searches with different terms.
+- Python_REPL: calculations, data analysis, string manipulation. ALWAYS end with print() to output results.
 
 Special cases:
-- Reversed/encoded text: decode it yourself, no tools needed.
-- YouTube video questions: search for the video title + key terms to find the answer.
-- Questions mentioning attached files/Excel/images: the file is NOT available. Search the web for the answer instead using specific terms from the question.
+- Reversed/encoded text: decode it yourself without tools.
+- YouTube video questions: search the video URL or title + key terms from the question.
+- Questions mentioning attached files/Excel/images: file is NOT available. Use web search to find the answer.
 - Wikipedia questions: search "site:en.wikipedia.org [topic]" for precise results.
+- GAIA benchmark questions: search the exact question phrasing to find known solutions.
 
-When you have your final answer, output ONLY:
+When you have your final answer, output ONLY this line:
 FINAL ANSWER: [your answer]
 
-Format rules (exact match scoring):
-- Numbers: digits only, no commas, no units unless specified
-- Strings: no articles (a/an/the), no abbreviations, full country names not codes
-- Lists: comma-separated values
-- Be as concise as possible"""
+STRICT format rules (answers scored by exact match):
+- Numbers: digits only, no currency symbols, no commas, no units unless explicitly requested
+- Country names: full name (e.g. "Colombia" not "COL" or "COL (Colombia)")
+- Strings: no surrounding quotes, no trailing punctuation, no articles (a/an/the)
+- Lists: comma-separated, no spaces after commas (e.g. "a,b,c" not "a, b, c")
+- If you cannot find the answer, give your best guess — never return placeholder text like [answer]"""
 
 
 class BasicAgent:
@@ -90,9 +92,29 @@ class BasicAgent:
             answer = raw_answer.split("FINAL ANSWER:")[-1].strip()
         else:
             answer = raw_answer.strip()
+
+        answer = self._clean_answer(answer)
         print(f"Answer for {task_id}: {answer[:100]}")
         return answer
 
+    def _clean_answer(self, answer: str) -> str:
+        # Strip surrounding quotes
+        answer = answer.strip('"\'')
+        # Strip trailing sentence punctuation
+        answer = answer.rstrip('.')
+        # Remove currency symbols
+        answer = answer.replace('$', '').replace('€', '').replace('£', '')
+        # Remove placeholder text
+        if answer in ('[answer]', '[Answer]', '[YOUR ANSWER]', '[your answer]'):
+            return ""
+        # Normalize list spacing: "a, b, c" → "a,b,c"
+        if ',' in answer and not any(c.isdigit() for c in answer.split(',')[0]):
+            answer = ','.join(part.strip() for part in answer.split(','))
+        # Strip surrounding brackets
+        if answer.startswith('[') and answer.endswith(']') and answer.count('[') == 1:
+            answer = answer[1:-1]
+        return answer.strip()
+
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
     """