fix: Separate zero-shot performance from few-shot

app.py

@@ -786,7 +786,8 @@ def fetch_results() -> dict[Language, pd.DataFrame]:
     logger.info("Fetching results from EuroEval benchmark...")
 
     response = requests.get(
-        "https://raw.githubusercontent.com/EuroEval/leaderboards/refs/heads/main/results/results.jsonl"
+        "https://raw.githubusercontent.com/EuroEval/leaderboards/refs/heads/main"
+        "/results/results.jsonl"
     )
     response.raise_for_status()
     records = [
@@ -805,6 +806,8 @@ def fetch_results() -> dict[Language, pd.DataFrame]:
         data_dict = defaultdict(dict)
         for record in records:
             model_name = record["model"]
+            if not record["few_shot"]:
+                model_name += " (zero-shot)"
             raw_results = record["results"]["raw"]
             if isinstance(raw_results, dict) and "test" in raw_results:
                 raw_results = raw_results.get("test", raw_results)