app: row-click detail panel below the leaderboard tables

Bundle 1+2 C6. Single shared `gr.Markdown` panel below both
Leaderboard widgets; either table's `.select(...)` event populates
it with the clicked row's notes, agent_url, submitted_at, submission
zip + report links (always), and failure_reason (failed rows only).
Spec'd shape: one panel, two select sources.

leaderboard.py:
- LEADERBOARD_COLS + VALIDATED_LEADERBOARD_COLS each grow by three
hidden detail-only fields (submission_id, notes, failure_reason).
They ride in the underlying DataFrame so the select handler can
read them out directly without a parallel state cache.
- New LEADERBOARD_HIDE_COLUMNS constant lists the columns hidden
from view; both Leaderboard widgets in app.py consume it.
- `_project_and_format` now fills missing columns with None (legacy
rows can lack optional fields like notes; the panel needs the
column-set to be stable regardless of source-row shape).

app.py:
- pandas imported (for the helper's pd.isna check).
- New `_format_detail(df, evt)` helper assembles the markdown:
pulls row by `evt.index` from the widget's currently-displayed
DataFrame; skips fields that are empty / null / NaN; preserves
the pre-formatted link cells (`[code](...)`, `[zip](...)`,
`[report](...)`) which markdown renders into clickable anchors.
Failure reason only shows on `status == "failed"` rows; report
link is already restricted to completed rows by leaderboard.py's
`_report_url_md`.
- Detail panel is a `gr.Markdown` initially carrying a placeholder
("Click a row above for details."). Both Leaderboards wire their
.select event to the same handler + the same output.

No new test file: the C6 acceptance is "row-click on either table
populates the detail panel; failed rows show failure_reason," which
is UI behavior verified on the live Space. Existing 9 unit tests
still pass since the column-list grows but the detail-only fields
are filled-or-None automatically by the new `_project_and_format`
fill-missing step. App.py import is clean, hidden cols are
present in the live DataFrame returned by load_leaderboard_split.

app.py

@@ -10,10 +10,13 @@ import logging
 import gradio as gr
 from gradio_leaderboard import Leaderboard
 
+import pandas as pd
+
 from leaderboard import (
     HF_DATA_REPO,
     HF_SUBMISSIONS_REPO,
     LEADERBOARD_DATATYPES,
+    LEADERBOARD_HIDE_COLUMNS,
     VALIDATED_LEADERBOARD_DATATYPES,
     load_leaderboard_split,
 )
@@ -45,6 +48,56 @@ correct 3D model.
 - **Code**: [`huggingface/cadgenbench`](https://github.com/huggingface/cadgenbench).
 """
 
+DETAIL_PLACEHOLDER = "_Click a row above for details._"
+
+
+def _has(value) -> bool:
+    """True for values that should show up in the detail panel."""
+    if value is None:
+        return False
+    if isinstance(value, float) and pd.isna(value):
+        return False
+    return str(value).strip() != ""
+
+
+def _format_detail(df: pd.DataFrame | None, evt: gr.SelectData) -> str:
+    """Build the row-detail markdown for the clicked submission.
+
+    Pulls hidden columns (``notes``, ``failure_reason``,
+    ``submission_id``) that ride in the underlying DataFrame plus
+    the visible link cells (already pre-formatted as ``[label](url)``
+    by ``leaderboard.py``'s ``_project_and_format``).
+    ``failure_reason`` only shows on ``failed`` rows;
+    ``report_url`` is only non-empty for ``completed`` rows.
+    """
+    if df is None or len(df) == 0 or evt is None or evt.index is None:
+        return DETAIL_PLACEHOLDER
+    idx = evt.index[0] if isinstance(evt.index, (list, tuple)) else evt.index
+    if idx < 0 or idx >= len(df):
+        return DETAIL_PLACEHOLDER
+    row = df.iloc[idx]
+
+    title = row.get("submission_name") or "(unnamed submission)"
+    lines = [f"### {title}", ""]
+    if _has(row.get("submitter_name")):
+        lines.append(f"- **Submitter**: {row['submitter_name']}")
+    if _has(row.get("status")):
+        lines.append(f"- **Status**: {row['status']}")
+    if _has(row.get("submitted_at")):
+        lines.append(f"- **Submitted**: {row['submitted_at']}")
+    if _has(row.get("notes")):
+        lines.append(f"- **Notes**: {row['notes']}")
+    if _has(row.get("agent_url")):
+        lines.append(f"- **Agent code**: {row['agent_url']}")
+    if _has(row.get("submission_blob_url")):
+        lines.append(f"- **Submission**: {row['submission_blob_url']}")
+    if _has(row.get("report_url")):
+        lines.append(f"- **Report**: {row['report_url']}")
+    if row.get("status") == "failed" and _has(row.get("failure_reason")):
+        lines.append(f"- **Failure reason**: {row['failure_reason']}")
+    return "\n".join(lines)
+
+
 with gr.Blocks(title="CADGenBench Leaderboard", theme=gr.themes.Soft()) as app:
     gr.Markdown(
         "# CADGenBench Leaderboard\n"
@@ -61,12 +114,14 @@ with gr.Blocks(title="CADGenBench Leaderboard", theme=gr.themes.Soft()) as app:
             value=initial_validated,
             datatype=VALIDATED_LEADERBOARD_DATATYPES,
             search_columns=["submission_name", "submitter_name"],
+            hide_columns=LEADERBOARD_HIDE_COLUMNS,
             label="Validated Leaderboard",
         )
         unvalidated_view = Leaderboard(
             value=initial_unvalidated,
             datatype=LEADERBOARD_DATATYPES,
             search_columns=["submission_name", "submitter_name"],
+            hide_columns=LEADERBOARD_HIDE_COLUMNS,
             label="Unvalidated Leaderboard",
         )
         refresh_btn = gr.Button("Refresh", size="sm")
@@ -75,6 +130,26 @@ with gr.Blocks(title="CADGenBench Leaderboard", theme=gr.themes.Soft()) as app:
             outputs=[validated_view, unvalidated_view],
         )
 
+        # Row-click detail panel: a single shared markdown component
+        # below the tables, populated by either table's .select()
+        # event. Pulls hidden columns (notes, failure_reason,
+        # submission_id) plus the visible link cells (already pre-
+        # formatted as markdown by leaderboard.py) for the row.
+        detail_panel = gr.Markdown(
+            value=DETAIL_PLACEHOLDER,
+            label="Selected submission",
+        )
+        validated_view.select(
+            fn=_format_detail,
+            inputs=validated_view,
+            outputs=detail_panel,
+        )
+        unvalidated_view.select(
+            fn=_format_detail,
+            inputs=unvalidated_view,
+            outputs=detail_panel,
+        )
+
     with gr.Tab("Submit"):
         gr.Markdown(
             f"""

leaderboard.py

@@ -29,6 +29,10 @@ LOCAL_RESULTS_PATH = Path(__file__).parent / "results.jsonl"
 RESULTS_FILENAME = "results.jsonl"
 HUB_FETCH_TIMEOUT_SECONDS = 30
 
+# Columns visible in the rendered table, in left-to-right order, followed
+# by hidden-but-data-present columns the row-click detail panel pulls from.
+# Hidden columns ride along in the DataFrame so `Leaderboard.select(...)`
+# can read them out without a separate state-cache or re-fetch.
 LEADERBOARD_COLS = [
     "status",
     "submission_name",
@@ -40,6 +44,10 @@ LEADERBOARD_COLS = [
     "agent_url",
     "submission_blob_url",
     "report_url",
+    # Detail-panel-only (hidden via `hide_columns` on the widget):
+    "submission_id",
+    "notes",
+    "failure_reason",
 ]
 
 # Validated table additionally exposes `validation_method`; on the
@@ -57,8 +65,16 @@ VALIDATED_LEADERBOARD_COLS = [
     "agent_url",
     "submission_blob_url",
     "report_url",
+    "submission_id",
+    "notes",
+    "failure_reason",
 ]
 
+# Columns to hide from rendering on both tables. These ride in the
+# DataFrame so the row-click detail panel can populate from them; the
+# widget hides them from view.
+LEADERBOARD_HIDE_COLUMNS = ["submission_id", "notes", "failure_reason"]
+
 # Per-column gradio_leaderboard datatypes. Link columns render their
 # pre-formatted markdown; everything else is plain string (numeric
 # cells get pending / failed status tags applied by _fmt_pct /
@@ -252,9 +268,15 @@ def _project_and_format(df: pd.DataFrame, columns: list[str]) -> pd.DataFrame:
         df["report_url"] = df.apply(
             lambda r: _report_url_md(r["submission_id"], r["status"]), axis=1,
         )
-    cols = [c for c in columns if c in df.columns]
+    # Make sure every declared column exists (legacy rows can be
+    # missing optional fields). Detail-panel reads expect the
+    # column-set to be stable regardless of which source rows had
+    # which keys.
+    for c in columns:
+        if c not in df.columns:
+            df[c] = None
     out = (
-        df[cols]
+        df[columns]
         .sort_values("aggregate_score", ascending=False, na_position="last")
         .reset_index(drop=True)
     )