GPTSwarm: Language Agents as Optimizable Graphs

Mingchen Zhuge, Wenyi Wang, Louis Kirsch, Francesco Faccio, Dmitrii Khizbullin, Jürgen Schmidhuber
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:62743-62767, 2024.

Abstract

Various human-designed prompt engineering techniques have been proposed to improve problem solvers based on Large Language Models (LLMs), yielding many disparate code bases. We unify these approaches by describing LLM-based agents as computational graphs. The nodes implement functions to process multimodal data or query LLMs, and the edges describe the information flow between operations. Graphs can be recursively combined into larger composite graphs representing hierarchies of inter-agent collaboration (where edges connect operations of different agents). Our novel automatic graph optimizers (1) refine node-level LLM prompts (node optimization) and (2) improve agent orchestration by changing graph connectivity (edge optimization). Experiments demonstrate that our framework can be used to efficiently develop, integrate, and automatically improve various LLM agents. Our code is public.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-zhuge24a, title = {{GPTS}warm: Language Agents as Optimizable Graphs}, author = {Zhuge, Mingchen and Wang, Wenyi and Kirsch, Louis and Faccio, Francesco and Khizbullin, Dmitrii and Schmidhuber, J\"{u}rgen}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {62743--62767}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/zhuge24a/zhuge24a.pdf}, url = {https://proceedings.mlr.press/v235/zhuge24a.html}, abstract = {Various human-designed prompt engineering techniques have been proposed to improve problem solvers based on Large Language Models (LLMs), yielding many disparate code bases. We unify these approaches by describing LLM-based agents as computational graphs. The nodes implement functions to process multimodal data or query LLMs, and the edges describe the information flow between operations. Graphs can be recursively combined into larger composite graphs representing hierarchies of inter-agent collaboration (where edges connect operations of different agents). Our novel automatic graph optimizers (1) refine node-level LLM prompts (node optimization) and (2) improve agent orchestration by changing graph connectivity (edge optimization). Experiments demonstrate that our framework can be used to efficiently develop, integrate, and automatically improve various LLM agents. Our code is public.} }
Endnote
%0 Conference Paper %T GPTSwarm: Language Agents as Optimizable Graphs %A Mingchen Zhuge %A Wenyi Wang %A Louis Kirsch %A Francesco Faccio %A Dmitrii Khizbullin %A Jürgen Schmidhuber %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-zhuge24a %I PMLR %P 62743--62767 %U https://proceedings.mlr.press/v235/zhuge24a.html %V 235 %X Various human-designed prompt engineering techniques have been proposed to improve problem solvers based on Large Language Models (LLMs), yielding many disparate code bases. We unify these approaches by describing LLM-based agents as computational graphs. The nodes implement functions to process multimodal data or query LLMs, and the edges describe the information flow between operations. Graphs can be recursively combined into larger composite graphs representing hierarchies of inter-agent collaboration (where edges connect operations of different agents). Our novel automatic graph optimizers (1) refine node-level LLM prompts (node optimization) and (2) improve agent orchestration by changing graph connectivity (edge optimization). Experiments demonstrate that our framework can be used to efficiently develop, integrate, and automatically improve various LLM agents. Our code is public.
APA
Zhuge, M., Wang, W., Kirsch, L., Faccio, F., Khizbullin, D. & Schmidhuber, J.. (2024). GPTSwarm: Language Agents as Optimizable Graphs. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:62743-62767 Available from https://proceedings.mlr.press/v235/zhuge24a.html.

Related Material