GraphSkillEvo: Evolutionary Optimization of Graph-Structured Agent Skills
GraphSkillEvo evolves graph-structured LLM agent skills with mutation and crossover, beating SkillOpt by 4.01% on GPT-5.4-nano.
The paper represents agent skills as graph-structured natural-language artifacts where nodes are execution steps with operational guidance and directed edges encode context-dependent transitions. GraphSkillEvo applies population-based evolutionary optimization with mutation and crossover operators over these graphs, exploring the structured skill space more broadly than iterative LLM self-refinement. Across five agent benchmarks it improves average accuracy by 4.01% on GPT-5.4-nano and 1.76% on GPT-5.4 over the SkillOpt baseline.
- Skills modeled as graphs: nodes are steps, edges are contextual transitions
- Population-based evolution with mutation and crossover operators
- Improves accuracy 4.01% (GPT-5.4-nano) and 1.76% (GPT-5.4) over SkillOpt
Full article210 words · extracted from huggingface.co · click to collapse
Skills can improve the performance of Large Language Model (LLM) agents by providing task-specific procedural guidance, while skill optimization further improves their effectiveness through iterative refinement. However, existing skill optimization methods typically represent skills as unstructured natural-language instructions, creating two key challenges: 1) Unstructured skills often lack explicit workflow-level guidance and contain substantial redundancy, making them difficult for LLMs to execute; 2) the vast search space of unconstrained natural-language skills makes skill optimization ineffective. To address these challenges, we propose representing skills as graph-structured natural-language artifacts. In graph-structured skills, each node represents an execution step together with its operational guidance, while directed edges encode context-dependent transitions between steps. Compared to unstructured skills, graph-structured skills can provide clear workflow-level guidance. Moreover, the proposed graph-structured skill can also facilitate skill optimization. Building on this structured representation, we introduce GraphSkillEvo, a population-based evolutionary optimization framework with mutation and crossover operators for graph-structured skills. By maintaining multiple candidate skills and combining effective components, GraphSkillEvo enables broader and more comprehensive exploration of the structured skill space than purely LLM-based iterative self-refinement. Extensive experiments across five agent benchmarks demonstrate that GraphSkillEvo consistently outperforms the strong skill optimization baseline SkillOpt, improving average accuracy by 4.01% on GPT-5.4-nano and 1.76% on GPT-5.4. Our code is available at https://github.com/ruisun7/GraphSkillEvo.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.21749