BONSAI: Evolvability-Guided Tree Search over Skills

2026-08-07

AI Summary

This work addresses the limitations of conventional skill optimization methods, which rely on a single reward metric and struggle to distinguish between overfitting spikes and sustainable plateaus, thereby hindering long-term evolutionary progress. To overcome this, the paper introduces an evolvability-guided skill optimization framework, pioneering the incorporation of the biological concept of “evolvability” into this domain. The approach employs Monte Carlo Tree Search to explore mutations in the skill textual space and devises an upper confidence bound (UCB) selection strategy that integrates both a node’s intrinsic score and the average fitness of its neighborhood. This enables dynamic allocation of optimization resources without incurring additional evaluation costs. Experiments demonstrate that the method improves zero-shot accuracy of a frozen 30B model by an average of 23.13 points across three benchmarks, significantly outperforming GEPA and SkillOpt baselines by 3.87 and 3.97 points, respectively.

Abstract

A skill is a naturallanguage document that steers a frozen agent whose weights cannot be updated so any capability the agent lacks must be supplied in prose Optimising a skill is therefore optimising text against a score and the standard recipe which keeps any edit that raises a heldout score is blind in a specific way a single score cannot tell a document perched on a narrow overfit spike from one resting on a broad plateau even though only the second can still be improved We introduce BONSAI a novel skilloptimisation framework that steers instead by evolvability the capacity of a region of documentspace to keep producing viable variation under further mutation a property biology treats as separate from present fitness BONSAI grows skills as a MonteCarlo search tree in which every child document is a mutation of its parent and descends it under an upperconfidence selection rule whose exploitation term blends a skills own fitness with the fitness of its mutational neighbourhood Because every child is a mutation the mean score recorded beneath a node estimates that neighbourhoods evolvability at no extra cost so the rule concentrates budget on regions that keep improving while its exploration term keeps a currently weak branch in contention BONSAI ships the single bestscoring document it finds at no cost beyond the acceptifbetter loop it replaces With a frozen 30B agent and averaged over three benchmarks BONSAI lifts heldout accuracy over the skillfree agent by 2313 points and improves on two budgetmatched baselines GEPA and SkillOpt by 387 and 397 points respectively

Problem

• skill optimization
• evolvability
• frozen agent
• natural language skills
• overfitting

Innovation

• evolvability
• skill optimization
• Monte Carlo tree search
• frozen language agent
• natural language skills