README: restore original voice, tasks overview covers both tasks, clarify countries and citations

Restructure README: Task 1 and Task 2 sections, GitHub starter kits; citations already complete

Task 2: single config with one merged jsonl per split

Task 2: one merged jsonl per split (captions + gold + image paths); drop baselines and per-file tsvs

Consolidated repo: rename, add Task 2 section and configs, mark all gold labels released

Add Task 2 (CRAI-Bench): train/dev/test with captions, human gold labels, reference and generated images, organizer baselines

Release test gold labels; add country/category/subcategory metadata to devtest and test

Add gold labels to devtest split

Declare test split in dataset configs

Release blind test split (task1a, task1b)

Add media (part 8)

Add media (part 7)

Add media (part 6)

Add media (part 5)

Add media (part 4)

Add media (part 3)

Add media (part 2)

Add media

updated bib

Update Slack invite link

Fix dataset card: restore task1b configs (was broken by stale task1c paths); keep Slack link

Remove em-dashes from dataset card

Dataset card: replace Google registration form with Slack channel

Codabench links -> re-uploaded comps (1a 17048/17049, 1b 17050/17051)

Task 1b: ranking metric -> Contrastive Instability; CFHR/Q+/Q- diagnostics; Codabench 17045/17046; greedy-decoding note

Update README.md

Update README.md

Update README.md

Update Task 1b Codabench links to new competitions (17022 en / 17021 msa)

Remove old task1c/ folder (renamed to task1b)

Add task1b/ data (copy of task1c) to match card configs

Update README.md

updated subtask id

Card: call the Fanar cascade an example, not a baseline (consistency with task pages)

Clean dataset card: remove em-dashes (no content or number changes)

ImageEval 2026 Task 1 (Ayn-VQA) dataset