Optimas: Optimizing Compound AI Systems with Globally Aligned Local Rewards
This addresses the problem of efficiently optimizing complex, multi-component AI systems for practitioners, though it appears incremental as it builds on existing optimization methods with a novel alignment approach.
The paper tackles the challenge of optimizing compound AI systems with non-differentiable structures and diverse configurations by proposing Optimas, a framework that uses locally aligned reward functions to enable independent component updates, resulting in an average performance improvement of 11.92% across five real-world systems.
Compound AI systems integrating multiple components, such as Large Language Models, specialized tools, and traditional machine learning models, are increasingly deployed to solve complex real-world tasks. However, optimizing compound systems remains challenging due to their non-differentiable structures and diverse configuration types across components, including prompts, hyperparameters, and model parameters. To address this challenge, we propose Optimas, a unified framework for effective optimization of compound systems. The core idea of Optimas is to maintain one Local Reward Function (LRF) per component, each satisfying a local-global alignment property, i.e., each component's local reward correlates with the global system performance. In each iteration, Optimas efficiently adapts the LRFs to maintain this property while simultaneously maximizing each component's local reward. This approach enables independent updates of heterogeneous configurations using the designated optimization method, while ensuring that local improvements consistently lead to performance gains. We present extensive evaluations across five real-world compound systems to demonstrate that Optimas outperforms strong baselines by an average improvement of 11.92%, offering a general and effective approach for improving compound systems. Our website is at https://optimas.stanford.edu.