Open-Ended Task Discovery via Bayesian Optimization

TL;DR

The paper introduces GSR, a Bayesian optimization framework that dynamically generates and refines tasks, improving optimization in open-ended scientific workflows with minimal regret overhead.

Contribution

It proposes a novel open-ended BO method that alternates between task generation and optimization, effectively handling evolving tasks in scientific applications.

Findings

GSR outperforms existing LLM-based optimizers in various applications.

It concentrates evaluations on the best task with logarithmic regret overhead.

Demonstrates effectiveness in product development, chemical synthesis, algorithm analysis, and patent repurposing.

Abstract

When applying Bayesian optimization (BO) to scientific workflow, a major yet often overlooked source of uncertainty is the task itself -- namely, what to optimize and how to evaluate it -- which can evolve as evidence accumulates. We introduce Generate-Select-Refine (GSR), a open-ended BO framework that alternates between task generation and task optimization. Starting from a user-provided seed task, GSR generates new tasks in a coarse-to-fine manner while a task-acquisition function schedules optimization. Asymptotically, it concentrates evaluations on the best task, incurring only logarithmic regret overhead relative to single-task BO. We apply GSR to new product development, chemical synthesis scaling, algorithm analysis, and patent repurposing, where it outperforms existing LLM-based optimizers.