These beautifully curated ‘mini-series’ are a unique and key feature of your work on this channel. And this series on graphs is one of the best, which is saying something. The series on sentence encoding/ Bert models was another one that comes to mind as something I go back to often.
Loved the video and wanted to go further. There seems to be a very similar technology called PLaG ( Plan Like a Graph) can you comment on the differences are they conflated? Thank you.
"PLaG includes a graph representation in the prompt, where we give models k-shot illustrations with graphs describing the task and instruct them to either reason based on a given graph (i.e., explicit graph) or to generate a graph themselves and then reason about it" - another method, quite different, published half a year ago. You find the link here: arxiv.org/pdf/2402.02805 "Graph-enhanced Large Language Models in Asynchronous Plan Reasoning" Code repo here: github.com/fangru-lin/graph-llm-asynchow-plan 51 stars
Recently and increasingly, AI is becoming a matter of national security in many places. If some countries have started to control the AI landscape trough regulations, why is China not making their researchers keep their papers as confidential? Not that they *should*, but I would expect by now that most countries were designating AI research as "classified", as militar and political strategy. And China and USA, with the current state of their relationship, shouldn't be already fully "locked down"?
These beautifully curated ‘mini-series’ are a unique and key feature of your work on this channel. And this series on graphs is one of the best, which is saying something. The series on sentence encoding/ Bert models was another one that comes to mind as something I go back to often.
I agree with you i love these videos
Wow, thank you!
I really appreciate the time you put into your videos! I feel more excited when I get back to working on my graphs after watching your content 😁
Great to hear!
Loved the video and wanted to go further. There seems to be a very similar technology called PLaG ( Plan Like a Graph) can you comment on the differences are they conflated? Thank you.
"PLaG includes a graph representation in the prompt, where we give models k-shot illustrations with graphs describing the task and instruct them to either reason based on a given graph (i.e., explicit graph) or to generate a graph themselves and then reason about it" - another method, quite different, published half a year ago.
You find the link here: arxiv.org/pdf/2402.02805
"Graph-enhanced Large Language Models in Asynchronous Plan Reasoning"
Code repo here:
github.com/fangru-lin/graph-llm-asynchow-plan
51 stars
I am actually working on putting a similar methodology in production! didn't even realize this paper is a thing, Thank you.
Glad it was helpful!
Does the AI say Tim Cook or Terry Gou?
Recently and increasingly, AI is becoming a matter of national security in many places. If some countries have started to control the AI landscape trough regulations, why is China not making their researchers keep their papers as confidential?
Not that they *should*, but I would expect by now that most countries were designating AI research as "classified", as militar and political strategy. And China and USA, with the current state of their relationship, shouldn't be already fully "locked down"?
Not last!
First! 🎉
I'm below so I'm first. Poser.
@ Noooo you beat me!