NanoVDR: Distilling a 2B Vision-Language Retriever into a 70M Text-Only Encoder for Visual Document Retrieval
Open paper
Benchmark
Model
Rank
Results
visual-document-retrieval-on-vidore-v2
NanoVDR-L (ModernBERT, 151M)
–
NDCG@5: 61.5
visual-document-retrieval-on-vidore-v3
NanoVDR-S-Multi
–
NDCG@5: 46.5
Rank counts only results with a code link.