MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval
Paper
•
2412.14475
•
Published
•
51
Multi-lingual Multi-domain Embedding Benchmark
community
AI & ML interests
None defined yet.
Recent Activity
View all activity
MMEB's activity
Shitao
authored
a
paper
10 days ago
nan
authored
4
papers
2 months ago
Jina Embeddings 2: 8192-Token General-Purpose Text Embeddings for Long Documents
Paper
•
2310.19923
•
Published
•
13
Multi-Task Contrastive Learning for 8192-Token Bilingual Text Embeddings
Paper
•
2402.17016
•
Published
•
5
Leveraging Passage Embeddings for Efficient Listwise Reranking with Large Language Models
Paper
•
2406.14848
•
Published
•
3
jina-embeddings-v3: Multilingual Embeddings With Task LoRA
Paper
•
2409.10173
•
Published
•
28
bwang0911
authored
4
papers
3 months ago
Leveraging Passage Embeddings for Efficient Listwise Reranking with Large Language Models
Paper
•
2406.14848
•
Published
•
3
Late Chunking: Contextual Chunk Embeddings Using Long-Context Embedding Models
Paper
•
2409.04701
•
Published
jina-embeddings-v3: Multilingual Embeddings With Task LoRA
Paper
•
2409.10173
•
Published
•
28
Jina-ColBERT-v2: A General-Purpose Multilingual Late Interaction Retriever
Paper
•
2408.16672
•
Published
•
7
Shitao
authored
a
paper
3 months ago
michael-guenther
authored
a
paper
3 months ago
zl101
authored
a
paper
4 months ago
michael-guenther
authored
a
paper
4 months ago
Post
2530
we are very proud to introduce
jinaai/jina-clip-v1, aka "jina-embeddings-multimodal".
The OpenAI CLIP openai/clip-vit-base-patch32 have nice performance to align text and image modality, that user can perform cross-modal text image retrieval or image classification on top of it. However, due to the training data and recipe, it can not:
1. model longer sequence of text inputs (77 token constraint).
2. align text representations (CLIP Text Tower is weak for text search).
In our latest publication, Jina CLIP: Your CLIP Model Is Also Your Text Retriever (2405.20204) , we proposed a multi-task, multi-objective learning scheme. The produced CLIP model shows:
1. Stronger cross-modal performance against OpenAI sets, 2% and 6% improvement on cross-modal retrieval recall@5.
2. Text tower of the JinaCLIP is a strong text encoder, reach the same performance as jinaai/jina-embeddings-v2-base-en, 165% improvement on MTEB[BEIR] recall@5.
3. Image tower of the JinaCLIP also shows strong performance in image-image search (CBIR), 12% recall improvement on Cifar100 test set.
If you are working on MuRAG (multimodal-retrieval argumented generation), try it out!
The OpenAI CLIP openai/clip-vit-base-patch32 have nice performance to align text and image modality, that user can perform cross-modal text image retrieval or image classification on top of it. However, due to the training data and recipe, it can not:
1. model longer sequence of text inputs (77 token constraint).
2. align text representations (CLIP Text Tower is weak for text search).
In our latest publication, Jina CLIP: Your CLIP Model Is Also Your Text Retriever (2405.20204) , we proposed a multi-task, multi-objective learning scheme. The produced CLIP model shows:
1. Stronger cross-modal performance against OpenAI sets, 2% and 6% improvement on cross-modal retrieval recall@5.
2. Text tower of the JinaCLIP is a strong text encoder, reach the same performance as jinaai/jina-embeddings-v2-base-en, 165% improvement on MTEB[BEIR] recall@5.
3. Image tower of the JinaCLIP also shows strong performance in image-image search (CBIR), 12% recall improvement on Cifar100 test set.
If you are working on MuRAG (multimodal-retrieval argumented generation), try it out!
nan
authored
a
paper
7 months ago
michael-guenther
authored
a
paper
7 months ago
bwang0911
authored
a
paper
7 months ago
Shitao
authored
a
paper
7 months ago
Shitao
authored
a
paper
8 months ago
Post
3053
In the vector search setup, we normally combine a fast embedding model and an accurate but slow reranker model.
The newly released @jinaai rerankers are small in size and almost as accurate as our base reranker. This means given a time constraint, it can scoring more candidate documents from embedding models and have a better chance to feed LLM the correct context for RAG generation.
These models are available on Huggingface and has been integrated into the latest SentenceTransformers 2.7.0. Check it out!
jinaai/jina-reranker-v1-turbo-en
jinaai/jina-reranker-v1-tiny-en
The newly released @jinaai rerankers are small in size and almost as accurate as our base reranker. This means given a time constraint, it can scoring more candidate documents from embedding models and have a better chance to feed LLM the correct context for RAG generation.
These models are available on Huggingface and has been integrated into the latest SentenceTransformers 2.7.0. Check it out!
jinaai/jina-reranker-v1-turbo-en
jinaai/jina-reranker-v1-tiny-en
- 1 reply