BLIP2

Reported on 4 benchmarks across 2 tasks · 2 papers · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Reasoning3 results

Visual ReasoningonWinoground
Group Score· 2023-05-10
19
best: 58.75 (GPT-4V (CoT, pick b/w two options))
Incorporating Structured Representations into Pretrained Vision & Language Models Using Scene Graphs arXiv:2305.06343
Visual ReasoningonWinoground
Image Score· 2023-05-10
23.8
best: 68.75 (GPT-4V (CoT, pick b/w two options))
Incorporating Structured Representations into Pretrained Vision & Language Models Using Scene Graphs arXiv:2305.06343
Visual ReasoningonWinoground
Text Score· 2023-05-10
42
best: 75.5 (GPT-4o + CA)
Incorporating Structured Representations into Pretrained Vision & Language Models Using Scene Graphs arXiv:2305.06343

Natural Language Processing1 result

Visual Question Answering (VQA)onInfoSeek
Accuracy· 2023-01-30
14.6
best: 30.65 (RA-VQAv2 w/ PreFLMR)
SOTA
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597