Jina AI

Wikimedia Deutschland e. V. startet in Kooperation mit den Experten von DataStax und Jina AI die Umsetzung eines semantischen Suchkonzepts, das die offen lizenzierten Daten von Wikidata für Entwickler*innen von KI-Anwendungen leichter nutzbar macht. Ziel ist es, den Prozess der Entwicklung von offenen, gemeinnützigen KI-Anwendungen zu vereinfachen und zu einem verlässlicheren Informations-Ökosystem beizutragen. Zur Pressemitteilung: https://lnkd.in/e7ZiyZth

We just released 𝐑𝐞𝐚𝐝𝐞𝐫-𝐋𝐌-𝟎.𝟓𝐁 and 𝐑𝐞𝐚𝐝𝐞𝐫-𝐋𝐌-𝟏.𝟓𝐁, two novel small language models (SLM) inspired by Jina Reader, designed to convert raw, noisy HTML from the open web into clean markdown. Both models are multilingual and support a context length of up to 𝟐𝟓𝟔𝐊 𝐭𝐨𝐤𝐞𝐧𝐬. Learn how we trained the model and how to use them. https://lnkd.in/escU9nQG

Jina ColBERT v2 supports 89 languages with superior retrieval performance, user-controlled output dimensions, and 8192 token-length. https://lnkd.in/eqygMDX8

You can't just use a CLIP model to retrieve text and images and sort the results by score. Why? Because of the modality gap. What is it, and where does it come from? https://lnkd.in/gRjgX2tm

Chunking long documents while preserving contextual information can be challenging. We introduce the "Late Chunking" strategy, which leverages long-context embedding models to generate contextual chunk embeddings, particularly useful for better retrieval applications. https://lnkd.in/dc68YzFs

What does it mean for LLMs when the web has been strip-mined clean, content providers have locked their doors, and there’s barely a trickle of new data to scrape? Let's dig in and find out. https://lnkd.in/ed-RuMpA

Jina Reranker v2, with its state-of-the-art multilingual performance, support for function calling and SQL querying, is now available on Azure Marketplace. This model listing offers enterprise users a secure and private deployment option directly within their Azure accounts. Built on top of the flexibility offered by Azure VMs, this listing enables deployment on Azure's generally available GPU instances. As a representative example of its performance, a 16GB T4 GPU VM offers up to 200x lower latency and over 30x higher concurrency compared to an Azure Container Instance (ACI) configuration with comparable CPU specs (GPU instances on ACI are not GA) — all at just double the instance pricing. Importantly, the fixed monthly price of Jina Reranker v2 remains unchanged over all supported VM instances, ensuring you get the most value out of your deployment. https://lnkd.in/diRp8aD8

We had a blast at ICML 2024 in Vienna, and we want to share with you everything we said, saw, and learned. https://lnkd.in/eq9mcr4c

When using embedding models for zero-shot classification, rephrasing the class label to "This is seriously about 'LABEL'" gives higher accuracy vs. using LABEL alone. But how, and why? https://lnkd.in/ggdeZqjM

A lot of LLMs can't figure out that 9.11 is actually smaller than 9.9. Can our embedding and reranker models do any better? https://lnkd.in/eG2MVnuK