Use the exact vidore/colpali_train_set test originals for doc1-4 (doc2 re-encoded from its source PNG to JPEG)

Restore the model-cards note (accidentally overwritten), keep trust_remote_code removed

Drop unnecessary trust_remote_code from the SentenceTransformer example

Explain why this exists and why you'd use it

Finalize card: drop MultiVectorEncoder from intro and the multi-vector/late-interaction tags, add SentenceTransformer usage example

Add document page images (vidore/colpali_train_set test split)

Add example documents (Tevatron OmniEmbed assets + DailyTalk speech clips)

initial commit