Visual Retrieval-Augmented Generation on Visual-RAG

66.04Score

CLIP-B

Updated 20d ago

Evaluation Results

Method	Links
CLIP-B 2026.05		66.04
Ours (Ovis2.5-2B Surrogate) 2026.05		65.11
VLM2Vec-V2.0 2026.05		61.63
OpenCLIP 2026.05		60.29
GT (Human-Annotated Image as Input) 2026.05		58.82
BGE-MLLM-S2 2026.05		55.75
ViLoMem 2026.01		54.6
PolarMem 2026.01		53.4
Zero-Shot 2026.05		52.67
ViLoMem 2026.01		50.8
Vanilla RAG 2026.01		50
PolarMem 2026.01		49.9
PolarMem 2026.01		48.6
QWEN2.5-VL-32B 2026.01		47.9
ViLoMem 2026.01		47.9
Vanilla RAG 2026.01		44.1
PolarMem 2026.01		44.1
PolarMem 2026.01		43.6
ViLoMem 2026.01		42.7
PolarMem 2026.01		42.6
ViLoMem 2026.01		42
ViLoMem 2026.01		41.8
PolarMem 2026.01		41.3
QWEN2.5-VL-7B 2026.01		40.2
ViLoMem 2026.01		40.2
InternVL-3.5-8B 2026.01		40.1
ViLoMem 2026.01		40
Vanilla RAG 2026.01		39.6
PolarMem 2026.01		39.5
Vanilla RAG 2026.01		39.4
DeepSeek-VL2 2026.01		38.2
Vanilla RAG 2026.01		38.1
Vanilla RAG 2026.01		37.9
InternVL-3.5-4B 2026.01		37.8
LLaVA-Next-Llama-3-8B 2026.01		36.8
Vanilla RAG 2026.01		35.9
Vanilla RAG 2026.01		35.6
DeepSeek-VL2-Small 2026.01		35.2
LLaVA-Next-Mistral-7B 2026.01		35.2