diff --git a/AGENTS.md b/AGENTS.md
index d2663f654983ecd64642835be8deb634c89258b3..a5b0c47a0ad743e508e13fdfa250bf4f8a847303 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -6,15 +6,16 @@ run no compute, so they are free; do not switch it to Gradio or Docker (those ne
## Files
-- `index.html`: the whole page: header, the five tab panels (Performance, Stability, Validity,
- Data overview, About), all user-facing text, the citation, ``
+- `index.html`: the whole page: header, the three tab panels (Performance, Stability, Validity),
+ all user-facing text (each explanation sits next to the figure or table it describes; there is no
+ About tab), the citation, ``
and the version history (`
`, newest first).
- `assets/app.js`: plain JavaScript, no build step. Fetches `data/meta.json`, renders the selector
chips (task type, downstream model, budget; combinations that are not published are disabled;
task type, view and noise level of the lasagna plots), sets the iframes' `src` and the figures'
`` to the generated files and resizes the iframes when a page posts `{saFrameHeight}`.
- Keeps the tab and the selections in the URL hash (old hashes `#leaderboard`, `#robustness` and
- `#datasets` still work).
+ Keeps the tab and the selections in the URL hash (old tab names `leaderboard`, `robustness`,
+ `datasets`, `data` and `about` still work).
- `assets/style.css`: color tokens on `:root` (light) and in `prefers-color-scheme: dark`.
- `smoke_test.py`: standard-library check of `data/` and of the version (see below).
@@ -32,7 +33,7 @@ run no compute, so they are free; do not switch it to Gradio or Docker (those ne
- `lasagna/stability/tasks_/` and `lasagna/validity/noise_/tasks_/`: one
`.svg` per entry of `lasagna.variants` and `per_epv.csv` (the plotted values).
- `results.csv` and `results_explorer.html`: the results per dataset, method, downstream model and
- budget (Data overview tab).
+ budget (Performance tab, below the leaderboard; not filtered by the chips).
- `per_dataset.csv`, `per_dataset_table.html` (datasets by name), and optionally `stability.csv`,
`stability_table.html`, `validity.csv` and one `validity_noise__table.html` per noise level
(listed in `meta.json` as `validity_tables`).
@@ -55,5 +56,5 @@ together.
(`
YYYY/MM/DD · vX.Y.Z: ...
`). `smoke_test.py` fails if they differ.
- No external scripts, fonts or CDNs; everything is served from this repo.
- Verification: `python smoke_test.py`, then `python -m http.server 8000` and a look at
- http://localhost:8000 (all five tabs, a few selections, light and dark mode).
+ http://localhost:8000 (all three tabs, a few selections, light and dark mode).
- Do not commit or push without an explicit request: pushing deploys the Space.
diff --git a/assets/app.js b/assets/app.js
index b2a11474278a2781f0338a6975ff6715d491b5e0..5fd80c7b6586a8ab941a5932cb26805f949c5add 100644
--- a/assets/app.js
+++ b/assets/app.js
@@ -22,9 +22,10 @@
overlay_top_3: 'Top-3 outlined',
binary_thr_003: 'Within 0.03 of the best'
};
- var TABS = ['performance', 'stability', 'validity', 'data', 'about'];
+ var TABS = ['performance', 'stability', 'validity'];
// Old links (#tab=leaderboard etc.) still open the matching tab.
- var TAB_ALIASES = { leaderboard: 'performance', robustness: 'stability', datasets: 'data' };
+ // Old tab names (links to earlier versions of the site).
+ var TAB_ALIASES = { leaderboard: 'performance', robustness: 'stability', datasets: 'performance', data: 'performance', about: 'performance' };
var DEFAULTS = {
tab: 'performance', tasks: ALL, model: ALL, budget: ALL,
stasks: ALL, svariant: 'plain', vtasks: ALL, vvariant: 'plain', noise: ''
@@ -142,7 +143,6 @@
});
if (tab === 'stability') renderStability();
if (tab === 'validity') renderValidity();
- if (tab === 'data') loadData();
writeHash();
}
function initTabs() {
@@ -301,14 +301,12 @@
showLasagna('validity', level.path, state.vtasks, state.vvariant);
}
- // ---- data overview ----
- function loadData() {
- var hasExplorer = !!(meta && meta.has_results_explorer);
- $('data-missing').hidden = !!meta;
+ // ---- results per dataset (Performance tab; not filtered by the chips) ----
+ function loadResults() {
+ var hasExplorer = !!meta.has_results_explorer;
$('frame-results').hidden = !hasExplorer;
if (hasExplorer) setFrame($('frame-results'), 'data/results_explorer.html');
- $('frame-datasets').hidden = !meta;
- if (meta) setFrame($('frame-datasets'), 'data/per_dataset_table.html');
+ setFrame($('frame-datasets'), 'data/per_dataset_table.html');
}
// ---- start ----
@@ -327,6 +325,7 @@
$('leaderboard-content').hidden = false;
renderFacts();
renderPerformance();
+ loadResults();
showTab(state.tab);
})
.catch(function () {
diff --git a/assets/style.css b/assets/style.css
index 14e384825edc3848a5913cd13f042b26284658db..6f76a9bc15e0c5e1039dfde8789a20aee2508596 100644
--- a/assets/style.css
+++ b/assets/style.css
@@ -106,6 +106,10 @@ a { color: var(--accent); }
h2 { font-size: 1.2rem; margin: 26px 0 8px; }
h3 { font-size: 1rem; margin: 20px 0 6px; color: var(--muted); }
.prose { max-width: 80ch; }
+.note { font-size: 0.85rem; color: var(--muted); max-width: 95ch; margin: 8px 0; }
+.caption p { margin: 0 0 6px; }
+.caption ul { margin: 0; padding-left: 20px; }
+.caption li { margin: 3px 0; }
.prose li { margin: 4px 0; }
code { font-size: 0.88em; background: var(--surface); padding: 1px 5px; border-radius: 4px; }
diff --git a/data/meta.json b/data/meta.json
index 954d52e86eb9208d37d9f0d22be545fe9f9492eb..6adce1d319122fba25200a5cd92c0e3635e6ea3f 100644
--- a/data/meta.json
+++ b/data/meta.json
@@ -1,6 +1,6 @@
{
"benchmark_name": "feature_selection_benchmark_2026_all",
- "generated_at": "2026-11-03 21:10 UTC",
+ "generated_at": "2026-11-03 21:45 UTC",
"selectarena_version": "0.1.0",
"n_results": 79873,
"n_datasets": 71,
diff --git a/data/subsets/tasks_all/models_all/budgets_0/elo_explorer.html b/data/subsets/tasks_all/models_all/budgets_0/elo_explorer.html
index f164f6ed2786d8bf4e0135eeff427a77372f4381..39f861ec9d4aadff62d163b026af60b4f88f196d 100644
--- a/data/subsets/tasks_all/models_all/budgets_0/elo_explorer.html
+++ b/data/subsets/tasks_all/models_all/budgets_0/elo_explorer.html
@@ -11,60 +11,20 @@ html, body { margin:0; padding:0; background:var(--bg); color:var(--fg);
.toolbar button { font:inherit; font-size:12px; color:var(--fg); background:var(--soft);
border:1px solid var(--line); border-radius:6px; padding:3px 10px; cursor:pointer; }
.toolbar button:hover { border-color:var(--accent); }
-