Learning to Configure Agentic AI Systems
Aditya Taparia, Som Sagar, Ransalu Senanayake
Why It Matters
What makes this one worth your time
This approach could optimize computational resources and enhance the accuracy of AI systems by tailoring configurations to specific queries, addressing the limitations of static configurations.
ARC dynamically configures LLM-based agents for improved performance.
Summary
The paper introduces ARC, a hierarchical policy for dynamically configuring LLM-based agent systems using a semi-Markov decision process, improving reasoning and tool-use accuracy by selecting query-specific configurations.
Key contributions
- Introduction of ARC, a hierarchical policy for agent configuration.
- Formulation of agent configuration as a semi-Markov decision process.
- Demonstrated improvements in reasoning and tool-use accuracy with query-specific configurations.
Notable insights
- Formulating agent configuration as a semi-Markov decision process allows for more flexible and adaptive AI systems.
- Using hierarchical policies to select configurations can significantly improve performance metrics across various benchmarks.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2602.11574v5 Announce Type: replace Abstract: Configuring LLM-based agent systems involves choosing workflows, tools, token budgets, and prompts from a large combinatorial design space, and is typically handled today by fixed templates or hand-tuned heuristics that apply the same configuration regardless of query difficulty, leading to brittle behavior and wasted compute. To address this, we formulate agent configuration as a semi-Markov decision process (SMDP) where each configuration acts as a temporally extended option that determines how an agent system processes a query, and introduce introduce ARC (Agentic Resource & Configuration learner), a lightweight hierarchical policy that dynamically selects query-specific agent configurations. Across reasoning, tool-use, and agentic benchmarks, ARC consistently improves over budget-matched tool-augmented LLMs, increasing average reasoning accuracy by 31.3%, tool-use accuracy by 13.95%, and doubling {\tau}-Bench (Airline) Pass^1 success from 9.0% to 18.0%. These results demonstrate that learning per-query agent configurations is a powerful alternative to "one size fits all" designs.