Skip to main content
PaperLLMsv1.0

BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models

by Salesforce Research · open-source · Last verified 2026-03-17

Presented BLIP-2, which bridges the modality gap between frozen image encoders and frozen LLMs using a lightweight Querying Transformer (Q-Former) trained in two stages. BLIP-2 achieves state-of-the-art VQA performance with significantly fewer trainable parameters than prior methods.

https://arxiv.org/abs/2301.12597
B+
B+Good
Adoption: AQuality: A+Freshness: B+Citations: AEngagement: F

Specifications

License
BSD-3-Clause
Pricing
open-source
Capabilities
visual-question-answering, image-captioning, image-text-retrieval, visual-reasoning
Integrations
huggingface
Use Cases
multimodal-qa, image-captioning, zero-shot-vqa
API Available
No
Tags
blip-2, multimodal, q-former, bootstrapping, vision-language
Added
2026-03-17
Completeness
100%

Index Score

71.9
Adoption
83
Quality
91
Freshness
78
Citations
82
Engagement
0

Put AI to work for your business

Deploy this paper alongside autonomous AaaS agents that handle tasks end-to-end — no babysitting required.

Explore the full AI ecosystem on Agents as a Service