Add Agent column

app.py

@@ -60,7 +60,12 @@ def init_leaderboard(dataframe, benchmark_type):
     if dataframe is None or dataframe.empty:
         raise ValueError("Leaderboard DataFrame is empty or None.")
     
-    AutoEvalColumnSubset = [c for c in fields(AutoEvalColumn) if ((c.name=="Model") or (TASK_NAME_INVERSE_MAP.get(c.name, dict()).get("type", "")==benchmark_type))]
+    non_task_cols = ["Model"]
+    if benchmark_type == "agentic":
+        # Add column for type of agent, as of now all models use the basic inspect agent
+        dataframe["Agent"] = ["[Basic Agent](https://inspect.ai-safety-institute.org.uk/agents.html#sec-basic-agent)"]*len(dataframe)
+        non_task_cols.append("Agent")
+    AutoEvalColumnSubset = [c for c in fields(AutoEvalColumn) if ((c.name in non_task_cols) or (TASK_NAME_INVERSE_MAP.get(c.name, dict()).get("type", "")==benchmark_type))]
 
     # styler = dataframe.style.apply(bold_max, subset=pd.IndexSlice[:, dataframe.columns[1:]])
 

src/display/utils.py

@@ -27,6 +27,7 @@ class ColumnContent:
 auto_eval_column_dict = []
 # Init
 auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
+auto_eval_column_dict.append(["agent", ColumnContent, ColumnContent("Agent", "markdown", True, never_hidden=True)])
 # Scores
 for task in Tasks:
     auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "markdown", True)])