Baidu: ERNIE 4.5 VL 424B A47B Model23/100 via “document understanding and information extraction from mixed-media content”
ERNIE-4.5-VL-424B-A47B is a multimodal Mixture-of-Experts (MoE) model from Baidu’s ERNIE 4.5 series, featuring 424B total parameters with 47B active per token. It is trained jointly on text and image data...
Unique: Combines visual layout understanding with semantic text extraction through MoE expert routing, where document structure experts handle spatial relationships and field localization while language experts perform semantic extraction. This dual-pathway approach avoids the brittleness of pure OCR or pure NLP approaches by leveraging both modalities.
vs others: More robust than OCR-only solutions for documents with complex layouts because it understands semantic context, while more efficient than dense vision-language models due to sparse expert activation for document-specific reasoning patterns.