Meta controller learns when ai agents should ask for help
Ask Your Advisor: Learning When to Consult Stronger Models in Agentic Systems
Multiagent Systems
Summary
Some AI systems use smaller models that try to solve tasks on their own and only ask bigger, smarter models when needed. But it's hard for the smaller model to know when its own thinking is wrong. The authors created a new way, ARIA, that learns when to ask for help by balancing success and cost. This method helps AI decide when to consult a stronger model, saving computation without losing much accuracy.
What this means in practice
- •For ai system builders: Implement AI agents that efficiently decide when to consult stronger models to reduce computation while maintaining performance.
- •For software infrastructure teams: Optimize backend resource use by applying ARIA to dynamically balance workload between light and heavy AI models in workflows.
Authors
Pascal Sowietzki, Steffen Holter, Yannick Metz, Emil Kasper, Raphael Baur, Mennatallah El-Assady
Abstract
To balance task performance and computational cost, agentic workflows can use a lower-cost model for execution and selectively consult a stronger model. In such executor-advisor systems, the executor decides when to seek advice, making escalation depend on self-assessment, although current language models struggle to detect where their own reasoning fails. The need for stronger reasoning emerges during execution and varies across workflow phases. We introduce ARIA, a meta-controller that learns when to consult by balancing task success against consultation cost. Our approach formulates consultation as a sequential resource-allocation problem. At workflow contact points, ARIA maps execution evidence to a compact state abstraction and selects SILENT or CONSULT, with action values learned by Monte Carlo control from cost-penalized outcomes. Across evaluated tasks of SWE-bench Verified, Terminal-Bench 2.1, and $τ^2$-bench, ARIA recovers 44-81% of the executor-advisor success gap at less than half the advisor model's cost. ARIA achieves the best observed success-cost trade-off among ten other evaluated conditions, including routing and cascading methods, fixed and prompted consultation, and random placement at matched advice volume. The policies are compact and inspectable, with a single consultation-price parameter governing the success-cost trade-off without retraining either underlying model.