Survey Study on AI Agent Architectures (2024)

by jslampe • 2 years ago

77 16 2 years ago

rck3• 2 years ago

Not mentioned in the paper, but I have been experimenting with behavior trees for LLM agents, and have had a lot of success: https://richardkelley.io/dendron/tutorial_intro/

muratsu• 2 years ago

This looks very interesting, thanks for sharing.

billmalarky• 2 years ago

Hi Richard I just reached out to you on LI. I'd love to get a chance to chat with you about your experience with this. Thank you for sharing this is fascinating.

gnat• 2 years ago

That’s a very well-written tutorial! I wish more software came with something so friendly and informative. Thanks.

rck3• 2 years ago

Thank you!

temporarely3• 2 years ago

How is a behavior tree different from a decision tree? Is there a subtle difference I am missing?

https://en.wikipedia.org/wiki/Decision_tree

sdesol• 2 years ago

This talks about the distinction

https://gamedev.stackexchange.com/questions/51693/difference...

rck3• 2 years ago

The other reply linked to a good explanation. I would only add that I also wrote a paper on Dendron, and Figure 17 shows how to transform a decision tree node into a behavior tree, so that you can implement any decision tree as a behavior tree:

https://arxiv.org/abs/2404.07439

abrichr• 2 years ago

Not mentioned: learning from demonstration. This is the approach we are taking at https://github.com/OpenAdaptAI/OpenAdapt.

sgt101• 2 years ago

I'll just draw folks attention to the long running AAMAS conference series.

Of course, it's quite academic in nature, but it may be that some useful approaches could be picked up from this resource for LLM driven approaches.

https://aamas2023.soton.ac.uk/

jrussino• 2 years ago

In addition, here are a few other conferences that may be of interest to people in this thread:

- AAAI: https://aaai.org/conference/aaai/

- ICAPS: https://icaps24.icaps-conference.org/

- IJCAI: https://www.ijcai.org/

RamblingCTO• 2 years ago

perfect timing! I'm just building myself an assistant via telegram and for now went with the multi-agent collaboration via supervisor pattern.

irthomasthomas• 2 years ago

"In the ever-evolving landscape of Natural Lan- guage Generation (NLG) evaluation, a noteworthy paradigm shift is underway as researchers increas- ingly turn their attention towards fine-tuning open- source language models (e.g., LLaMA), in lieu of traditional closed-based LLMs like ChatGPT and GPT-4. This transformative shift is propelled by a thorough exploration of key perspectives, including the expenses associated with API calls, the robust- ness of prompting, and the pivotal consideration of domain adaptability."

This paper was written by an LLM. Probably Claude-3.

Havoc• 2 years ago

Finding it quite difficult to decide which platform to bet on. Autogen langchain and langgraph seem to be main contenders. And then people seem to custom roll them too

Survey Study on AI Agent Architectures (2024)

Related Stories