Strategist evolves reasoning strategies for better AI task performance
Strategist: Evolving Reasoning with Agents
Artificial IntelligenceComputation and Language
Summary
AI systems use models to understand tasks, but they also need ways to decide how to use these models to give answers. Usually, people hand-write these reasoning methods each time, which is slow and doesn’t improve over time. The authors created Strategist, a kind of AI that writes and improves its own reasoning plans by reusing and refining strategies it learned from earlier tasks. This approach works well on many challenges and can even adapt to new tasks without extra work.
What this means in practice
- •For ai development teams: Automatically generate and improve reasoning methods across diverse AI tasks to reduce manual coding and speed up deployment.
- •For software engineers: Integrate evolving inference strategies to reduce computational cost while maintaining accuracy in AI-powered applications.
Authors
Taiming Lu, Qi Xu, Zhigang Hua, Zhuang Liu
Abstract
A model is only half of an AI agent system. The other half is the inference strategy: the reasoning procedure that turns model calls into a response. Today that second half is written by hand, and the experience of building one strategy never carries to the next. We introduce Strategist, a meta-agent that evolves inference strategies across tasks. Each strategy is written as a blueprint,executable code that other strategies can reuse, and together the blueprints form a library. Given a task and a small development set, Strategist retrieves related blueprints, writes candidate strategies, tests them on real samples, and revises them from the results. Every candidate then enters the library, so later tasks build on earlier discoveries. On 30 benchmarks, Strategist beats the strongest baseline on 24 and matches its average accuracy (74.7% vs.74.0%) at 39% lower inference cost. Its strategies also transfer to unseen tasks and other models without further search. The part of an AI system that is not improved on its own now does.