Problem
Research questions and friction points this paper is trying to address.
skill learning
agent cost
cost effectiveness
programmatic skills
LLM agents
Better, Faster, Stronger: Programmatic Skill Learning Best Reduces Agent Cost
🤖 AI Summary
This work proposes a procedural skill learning paradigm to reduce the computational and trial-and-error costs incurred by large language model agents when adapting to new domains. By modeling skills as executable programs, the approach automatically synthesizes and refactors efficient skills through trajectory analysis, enabling incremental reuse during inference without requiring replay or validation. Integrating deterministic action sequence execution with zero-shot skill transfer, the method substantially improves sample efficiency and reduces resource consumption across three embodied environments, while maintaining robustness to distributional shifts and environmental stochasticity. This study provides the first systematic demonstration of the cost-effectiveness advantages of procedural skill representations in agent learning.
📝 Abstract
Recently, the practice of augmenting LLM agent capability with skills has gained prevalence. We explore the cost effective adaptation of agents to novel domains by means of learning skills. Existing works focus on performance gain over cost effectiveness. As a result, little is known about what skill learning strategies save cost. We argue that among all the different skill learning methods, those that view skills as programs can achieve the best cost reduction. By executing sequences of actions deterministically, a program-augmented agent can reliably and cheaply achieve goals that would otherwise require trial and error and risk degenerate behavior over long horizons. An agent can learn at inference time by incrementally discovering these programs and equipping them for future tasks. We hypothesize that past trajectories contain enough signal to guide skill learning, even without replay or validation, provided the agent can learn to analyze them. To test our claims, we propose SpeedRunner, a coding agent that analyzes trajectories and refactors skills for better performance on future tasks. Across three different embodied environments, we show that SpeedRunner consistently achieves the frontier in learning and cost reduction while remaining robust against distribution shifts and environmental randomness.
Innovation
Methods, ideas, or system contributions that make the work stand out.
programmatic skills
cost-effective learning
trajectory analysis
inference-time skill acquisition
embodied agents
🔎 Similar Papers
No similar papers found.
