Commit History
-
leaderboard: rebuild Tasks tab as grouped thumbnail grid
Michael Rabinovich +1: CursorReplace the flat 84-row Tasks list with a responsive card grid grouped by type (Generation / Editing) with live count badges, a number search, and an All/Generation/Editing segmented filter. Each card shows the task's input geometry thumbnail (generation -> input.png; editing -> renders/iso.png, the iso render of input.step) plus the sample number and type tag. Thumbnails use loading="lazy" + decoding="async" with reserved 4:3 dimensions and CSS-constrained width so the 84 images don't all fetch at once. Card click opens the task's existing detail card (prompt + input), keeping j/k/Esc navigation. System fonts only. Co-authored-by: Cursor <cursoragent@cursor.com>95f3ee8 -
leaderboard: rename tabs, relabel fixtures as samples, inline gallery row stats
Michael Rabinovich +1: Cursor- Rename the visual "Gallery" tab to "Leaderboard" and the table tab to "Detailed View" (order unchanged); update the smoke test accordingly. - Replace user-facing "fixture" wording with "sample" across the gallery, task browser, submit instructions, and submitter status/error messages. Internal data keys, routes, CSS, and identifiers are unchanged. - Drop the gallery's "more numbers" expander: surface Generation/Editing inline under the aggregate score and move Download ZIP under the name. - Link each gallery submission title to its hosted report (same /reports/<id>.html the Detailed View table uses); drop the date line. Co-authored-by: Cursor <cursoragent@cursor.com>461547b
-
leaderboard: add Tasks tab to browse benchmark fixtures
Michael Rabinovich +1: CursorA read-only task browser that mirrors the per-submission report's summary-table -> detail-card navigation (j/k, Esc, Prev/Next, deep-linkable) but shows only the prompt + input (drawing for generation, starting-shape renders for editing) with no scores or ground truth. Fixtures come from each <fixture>/description.yaml in the inputs dataset; input images are proxied through the Space's read token via /task-input and lazy-loaded. Co-authored-by: Cursor <cursoragent@cursor.com>f4924d6