Evoskill v2 improves task-solving skills by evolving instructions automatically
EvoSkill v2: Fully Agentic Skill Evolution from Benchmark Adaptation to a Stateful Evolver
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
Some computer programs solve tasks by following instructions called skills. The authors created EvoSkill v2, a system that automatically improves these skills without changing the program itself. It works with three agents: one adapts the task setup, one tries to solve the task, and one evolves the skills by rewriting instructions based on past attempts. EvoSkill v2 showed better success rates than similar methods on two testing benchmarks and can even help less capable models improve their own skills.
What this means in practice
- •For automation engineers: Enable automated improvement of task instructions to boost robotic or software agent performance on benchmark problems.
- •For banking ai teams: Improve performance of AI models solving banking domain tasks by evolving skills using the described system.
Authors
Dastin Huang, Baran Nama, Salaheddin Alzubi
Abstract
EvoSkill v2 is a three-agent system that edits the instructions (skills) an agent is given, leaving its model and harness intact. The user starts a run by describing a benchmark and supplying a grader, which a benchmark adaptor turns into a benchmark-specific plugin. A solver attempts the tasks, and an evolver reads the solver's training rollouts and rewrites the skills at each epoch. We implement the evolver as one stateful coding agent with a bash tool for the whole run, unlike the two most common designs among the skill optimizers we surveyed: a sequence of single model calls, as in SkillOpt, and an agent that starts afresh each round, as in SkillBoost. We also measure how much the solver's rollouts vary from one attempt to the next, give the evolver several attempts per training task, and accept every epoch's skills instead of selecting among them on noisy scores. With fine-grained ground truth, EvoSkill v2's skills raised the held-out pass rate above the stronger of SkillOpt and SkillBoost by 7.7 percentage points on SpreadsheetBench v2 and by 5.4 on $τ^3$-bench banking. In both benchmarks and both feedback settings, EvoSkill v2's evolver cost less per percentage point of held-out pass rate gained than either baseline's optimizer. EvoSkill v2 also enables a less capable model to improve its own skills, demonstrated by a gain of up to 21.5 percentage points in the $τ^3$-bench banking experiment where Claude Haiku 5.5 powers both solver and evolver.