- Deduplicate GT by query_en to handle duplicate entries gracefully
- Tolerate up to 3 extraneous queries without triggering partial submission
- Prevents false 'incomplete submission' when GT has duplicates or
submission contains a few extra queries
Update nav to scheme C: Raw Images: obox · Google Drive
- Make refresh button a 60x60px rounded square (border-radius: 12px)
- Replace single Raw Images link with OneBox (pwd: Oppo2026) + Google Drive
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Fix navigation link, evaluation metrics, refresh alignment and tooltip locale
- Fix Protected Leaderboard URL in navigation
- Add k=100 to supported evaluation metrics
- Align refresh button with dropdown controls via CSS
- Inject script to clean ag-grid title attributes and hide CN tooltips
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- PhotoBench-Full (raw version) is now referred to as PhotoBench
- Original protected version is now PhotoBench-Protected
- Update about.py, app.py, README.md, and construct/guideline.md
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add link to PhotoBench-Full-HF dataset download page
- Add opt-in toggle for publishing results to public leaderboard
- Only full submissions (all 3 albums, complete queries) appear on leaderboard
- Add admin tab with password-protected panel for managing entries
- Add column selector to let users choose which metrics to display
- Default display metrics: R@1/5/20/50 and N@1/5/20/50
- All metrics R@1/5/10/20/50/100 and N@1/5/10/20/50/100 computed during evaluation
- Enlarged UI fonts to 17px base
- Remove evaluated_queries from leaderboard display
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
v2.1: remove Chinese text, Apple-style UI, centered submit form, auto-seed leaderboard data