Actually the best way to get support at Google is to start working on the product. Writing an emnlp paper might be a close second though.

I do earnestly think that page (and character) limits lead to better quality writing. I don't like arxiv 40 pagers or twitter blue (or whatever it's called) essays

Introducing Gecko 🦎, a new text embedding model from Google DeepMind! Distilled from LLMs, Gecko offers powerful embeddings for various NLP tasks. Gecko is now available in Google Cloud API 👉bit.ly/google-gecko-a… Paper: bit.ly/google-gecko Colab: bit.ly/google-gecko-c…

Gecko 🦎 is like promptagator🐊 in that it's also a reptile, but different in that it's a single model that changes uh... colors to blend in with its uh... surrounding task. Other important differences include being available through the google cloud api and me being on the paper

Google presents Can Long-Context Language Models Subsume Retrieval, RAG, SQL, and More? Long-context LM: - Often rivals SotA retrieval and RAG systems - But still struggles with areas like compositional reasoning repo: github.com/google-deepmin… abs: arxiv.org/abs/2406.13121

Can long-context language models (LCLMs) subsume retrieval, RAG, SQL, and more? Introducing LOFT: a benchmark stress-testing LCLMs on million-token tasks like retrieval, RAG, and SQL. Surprisingly, LCLMs rival specialized models trained for these tasks! arxiv.org/abs/2406.13121

"just put the corpus into the context"! Long context models can already match or beat various bespoke pipelines and infra in accuracy on non-trivial tasks! Hadn't expected this so soon, and honestly was hoping to milk RAG impact for a little longer 🤪

Excited to present a new benchmark "LOFT" to study long-context language models ability to do in-context retrieval, RAG, and complex reasoning tasks. We find that long-context models are getting increasingly capable and often rival task-specific experts for 1M context length.

Do long-context LMs obsolete retrieval, RAG, SQL and more? Excited to share our answer! arxiv.org/abs/2406.13121 from the team at Google DeepMind that wrote one of the 1st papers on RAG (REALM) and repeat SOTA on retrieval (Promptagator, Gecko). w/ Gemini 1.5 Pro, the answer is 🧵

Ever wondered if long-context language models can also master image, video, and multimodal retrieval? 🌟 Dive into our latest work LOFT! We benchmarked various long-context language models on million-token level retrieval, RAG, and SQL tasks across text, vision, and audio 🚀 #AI

Can long-context models replace retrievers, RAG & SQL? We evaluate them on smaller-scale versions of these tasks and compare them to specialized models in same settings. We found *prompting* LLM perform surprisingly well, generalizing across text, multimodal & other settings!

This is clearly making fun of UK food, before anyone from there gets too excited

Excited to launch and share the new citations feature ("related content") in Gemini! Last fall, we introduced Double-Check (shorturl.at/kFJqF), which checks Gemini's claims against sources on the web. With citations, we're now auto-running on all fact-seeking queries.