Google Research’s Post

Graphs, structures that describe connections between objects, are everywhere — imagine the tools in a kitchen, parts of a bike, or a group of friends. Learn about our latest work that explores how to encode graphs in a format that an LLM can understand: → https://goo.gle/3TbowPr

  • Pictured, the process of encoding a graph as text using two different approaches and feeding the text and a question about the graph to the LLM.
Federico Minutoli

Staff Research Engineer at Reply | OSS builder

1y

There's so much data out there which can be structured, and which organizations do structure, as graphs with similar visual properties enabling LLM processing, but still a long long way to go. ⛽

Very exciting! The interplay between graphs and LLMs are super interesting! Graphs are a super powerful representation and very efficient at inference time. However, they didn't manage to achieve their full potential, because it was so labor intensive to create them. Now with LLMs, there are new approaches to create graphs with relatively little manual work, and with this news - also use LLMs to significantly increase the value that can be extracted from the graph. Unlocks interesting use cases!

Dylan G.

Attention Is All We Need | Following: 🤗🪿🦉

1y

There is a bunch here to absorb. I think that the potential gains are significant, as so many organizational charts and process flows adopt this visual format. It is also desirable for many use cases for LLMs to generate graphs and visually structured information.

This should become Data101, replacing SQL.

Wallace Wong

Agency Owner at Farmers Insurance | Commercial / Business Insurance Specialist

8mo

Great & Useful!

Fo So

Web developer, Tech guy and startuper. Follow for more: 🚀

10mo

One of my preferred tool ✅

Like
Reply
Reza Ardestani

Data Scientist | Chatbot and Language Models Developer | Reinforcement Learning & Explainable AI Researcher

1y

Interesting.

See more comments

To view or add a comment, sign in

Explore topics