Scaling Small Agents Through Strategy Auctions

Lisa Alazraki, William F. Shen, Yoram Bachrach, Akhil Mathur
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:1706-1766, 2026.

Abstract

Small language models are viewed as a promising, cost-effective approach to agentic AI, yet how their performance scales with task complexity remains unclear. While smaller agents match larger ones on simple tasks, it is unknown when large models become necessary and how to better leverage small agents. In this work, we show that small agents fail to scale with task complexity on deep search and coding tasks, and introduce Strategy Auctions for Workload Efficiency (SALE), a framework inspired by freelancer marketplaces. In SALE, agents bid with strategic plans scored by a cost–value mechanism and refined via shared auction memory, enabling per-task routing and continual self-improvement without training a router. On average, SALE reduces reliance on the largest agent by 52%, lowers overall cost by 35%, and consistently improves pass@1 with only a negligible token overhead. In contrast, established routers either underperform the largest agent or fail to reduce cost. These results suggest that small agents can be effectively “scaled up” through coordinated allocation and test-time self-improvement. More broadly, they motivate a systems-level view of agentic AI in which gains come less from ever-larger individual models and more from market-inspired coordination mechanisms that organize heterogeneous agents into efficient, adaptive ecosystems.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-alazraki26a, title = {Scaling Small Agents Through Strategy Auctions}, author = {Alazraki, Lisa and Shen, William F. and Bachrach, Yoram and Mathur, Akhil}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {1706--1766}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/alazraki26a/alazraki26a.pdf}, url = {https://proceedings.mlr.press/v306/alazraki26a.html}, abstract = {Small language models are viewed as a promising, cost-effective approach to agentic AI, yet how their performance scales with task complexity remains unclear. While smaller agents match larger ones on simple tasks, it is unknown when large models become necessary and how to better leverage small agents. In this work, we show that small agents fail to scale with task complexity on deep search and coding tasks, and introduce Strategy Auctions for Workload Efficiency (SALE), a framework inspired by freelancer marketplaces. In SALE, agents bid with strategic plans scored by a cost–value mechanism and refined via shared auction memory, enabling per-task routing and continual self-improvement without training a router. On average, SALE reduces reliance on the largest agent by 52%, lowers overall cost by 35%, and consistently improves pass@1 with only a negligible token overhead. In contrast, established routers either underperform the largest agent or fail to reduce cost. These results suggest that small agents can be effectively “scaled up” through coordinated allocation and test-time self-improvement. More broadly, they motivate a systems-level view of agentic AI in which gains come less from ever-larger individual models and more from market-inspired coordination mechanisms that organize heterogeneous agents into efficient, adaptive ecosystems.} }
Endnote
%0 Conference Paper %T Scaling Small Agents Through Strategy Auctions %A Lisa Alazraki %A William F. Shen %A Yoram Bachrach %A Akhil Mathur %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-alazraki26a %I PMLR %P 1706--1766 %U https://proceedings.mlr.press/v306/alazraki26a.html %V 306 %X Small language models are viewed as a promising, cost-effective approach to agentic AI, yet how their performance scales with task complexity remains unclear. While smaller agents match larger ones on simple tasks, it is unknown when large models become necessary and how to better leverage small agents. In this work, we show that small agents fail to scale with task complexity on deep search and coding tasks, and introduce Strategy Auctions for Workload Efficiency (SALE), a framework inspired by freelancer marketplaces. In SALE, agents bid with strategic plans scored by a cost–value mechanism and refined via shared auction memory, enabling per-task routing and continual self-improvement without training a router. On average, SALE reduces reliance on the largest agent by 52%, lowers overall cost by 35%, and consistently improves pass@1 with only a negligible token overhead. In contrast, established routers either underperform the largest agent or fail to reduce cost. These results suggest that small agents can be effectively “scaled up” through coordinated allocation and test-time self-improvement. More broadly, they motivate a systems-level view of agentic AI in which gains come less from ever-larger individual models and more from market-inspired coordination mechanisms that organize heterogeneous agents into efficient, adaptive ecosystems.
APA
Alazraki, L., Shen, W.F., Bachrach, Y. & Mathur, A.. (2026). Scaling Small Agents Through Strategy Auctions. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:1706-1766 Available from https://proceedings.mlr.press/v306/alazraki26a.html.

Related Material