Han Xiao - Jina AI

Jina AI

Sign in Subscribe

Han Xiao

Founder & CEO of Jina AI

Diagram illustrating the 'Late Chunking' and 'Long Document Model' processes in machine learning on a black background.

Tech Blog Featured

Late Chunking in Long-Context Embedding Models

Chunking long documents while preserving contextual information is challenging. We introduce the "Late Chunking" that leverages long-context embedding models to generate contextual chunk embeddings for better retrieval applications.

Black background with a green pixel dinosaur in the center, surrounded by green and yellow text related to data labeling.

Rephrased Labels Improve Zero-Shot Text Classification by 30%

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?

Rows of numbered wooden pieces on a white background, ranging from single digits to high numbers.

Can Embedding/Reranker Models Compare Numbers?

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?

Abstract background with dynamic green particles and lines on a black backdrop, emitting a sense of motion and energy.

No. You Can't Use Reranker to Improve SEO

But if you work in SEO, it could be interesting to see things from the other side of the table; understand how embeddings and rerankers play their roles in modern search systems.

Cartoon of four characters in a cemetery with graves marked "RAG," mixing somber themes with humorous actions.

RAG is Dead, Again?

RAG is just one algorithmic pattern you can use. But if you make it *the* algorithm and idolize it, then you are living in a bubble you created, and the bubble will burst.

Airbnb CEO Brian Chesky and another executive smiling at a tech conference, surrounded by attendees.

What's Interesting in ICLR2024

With nearly 6000 in-person attendees, ICLR 2024 was easily the best and largest AI conference I've attended recently! Join me as I share my top picks—both the cherries and lemons—of prompt-related and model-related work from those top AI researchers.

Screenshot of a Tetris-like game with "Score: 40" and "Press Start 2P" text on display.

Tech Blog Featured

DSPy: Not Your Average Prompt Engineering

Heads up, Bay Area guys ditched their AVP already and buzz about DSPy now. Could DSPy be the new go-to framework for prompt engineering after LangChain and LlamaIndex?

Neon theater or concert hall marquee letters lit up at night with city lights and faint "Adobe Sto" visible.

Tech Blog Featured

What is ColBERT and Late Interaction and Why They Matter in Search?

Jina AI's ColBERT on Hugging Face has set Twitter abuzz, bringing a fresh perspective to search with its 8192-token capability. This article unpacks the nuances of ColBERT and ColBERTv2, showcasing their innovative designs and why their late interaction feature is a game-changer for search.

Diagram illustrating a neural network process with smiley faces and repeated mentions of "Similar" on a blackboard-like backg

Tech Blog Featured

Does Subspace Cosine Similarity Imply High-Dimensional Cosine Similarity?

Does high similarity in subspace assure a high overall similarity between vectors? This post examines the theory and practical implications of subspace similarity.

Illuminated sign reading "EMNLP 2023 Entry" mounted above a door, suggesting a conference entrance

A Tale of Two Worlds: EMNLP 2023 at Sentosa

Just back from EMNLP2023 and my mind's still reeling! Witnessed NLP's seismic shift firsthand through daring papers and provocative posters that are challenging everything we thought we knew. Check out my take on the conference's boldest ideas.

Conference attendees engage in discussion in front of a presentation screen displaying "DeepDive 2B" with names and affiliati

Give and Take in FDP's AI Geopolitics Panel

This blog post captures my personal reflections and contributions to the panel, highlighting key differences in AI strategies among global powers like the US, China, and the EU.

Illustration of a woman observing a shop display with jars under a red and black ambiance

Do You Truly Need a Dedicated Vector Store?

Unraveling vector search spaghetti: Lucene's charm vs. shiny vector stores. Navigating enterprise mazes & startup vibes. Where's search headed next?

Digital artwork of a human eye with rainbow colors and long black eyelashes in a surreal style

Tech Blog Featured

SceneXplain vs. MiniGPT4: A Comprehensive Benchmark of Top 5 Image Captioning Algorithms for Understanding Complex Scenes

Uncover the future of image captioning as SceneXplain and its rivals face off in an epic showdown. Explore their impact on accessibility, SEO, and storytelling, and dive into our intriguing results to witness the cutting-edge capabilities of these algorithms.

Five stylized heads in a row, labeled A to E, interconnected with lines and surrounded by colorful digital symbols and formul

Insights Featured

Auto-GPT Unmasked: The Hype and Hard Truths of Its Production Pitfalls

Auto-GPT: A groundbreaking project or just another overhyped AI experiment? In this blog post, we unravel the truth behind the buzz and reveal the limitations that make this project unsuitable for real-world applications.

Magical scene with a fairy in a green and gold dress reading a book under a tree, surrounded by butterflies

Tech Blog Featured

SceneXplain: Unleash the Advanced Image Captioning & Storytelling

Uncover the game-changing potential of SceneXplain, an advanced image captioning solution powered by LLMs. Check out the benchmark against Midjourney, CLIP, BLIP2, and other alternatives. Dive into our blog post and experience the revolution firsthand!

Mysterious figure in a black robe sitting on a leather chair, holding a white cup, with a plain dark background

Insights Featured

SEO is Dead, Long Live LLMO

With ChatGPT replacing Google search, traditional SEO is dead: how can businesses ensure their information appears in ChatGPT's single answer? The solution is in-context learning on LLMs.

Broken yellow piggy bank on a blue background with scattered coins around it, suggesting financial loss

Fine-tuning with Low Budget and High Expectations

Fine-tuning is highly effective, economical, and eco-friendly; or at least that's what all the deep learning tweets will tell you. Exactly how much data & time do you need to get a good result?

Abstract artwork with a figure looking up towards a cloudy sky, accented by striking contrasts and vivid colors

Insights Featured

Why Don't People Use Your Software Framework?

The initial hype was great: You got hundreds of GitHub stars and fifty upvotes on Reddit. But then growth started to plateau and then drop. Why don't people use your software framework?

Colorful laptop screen display with search and create buttons, icons, and the caption "Magic item that looks shiny.

Insights Featured

Search is Overfitted Create; Create is Underfitted Search

The biggest competition of neural search may come from a technology that doesn't even need embeddings as an intermediate representation – an end-to-end technology that directly returns the result you want.

Group of students posing together at the COLING 2022 event, projecting a sense of camaraderie

COLING2022 Summary on Multimodal AI

Last week I was in Gyeongju, Korea for the COLING2022 conference. COLING is the premier conference in computational linguistics held every two years. In this article, I will review the multimodal AI-related work in COLING 2022.

Vibrant image of an astronaut with a background of colorful light stripes and text reading "Multimodal AI, coming your way.

Insights Featured

The Paradigm Shift Towards Multimodal AI

We are on the cusp of a new era in AI, one in which multimodal AI will be the norm. At Jina AI, our MLOps platform helps businesses and developers win while they're right at the starting line of this paradigm shift, and build the applications of the future today.