CASCADE: Cumulative Agentic Skill Creation through Autonomous Development and Evolution

TL;DR

CASCADE is a self-evolving LLM agent framework that acquires and refines skills through autonomous development, significantly improving performance on complex scientific tasks and enabling scalable AI-assisted research.

Contribution

Introduces CASCADE, a novel framework for autonomous skill acquisition and evolution in LLM agents, advancing beyond static tool use to continuous learning and self-reflection.

Findings

Achieves 93.3% success rate on SciSkillBench with GPT-5.

Outperforms non-evolving agents by a large margin.

Demonstrates real-world scientific applications and knowledge sharing.

Abstract

Large language model (LLM) agents currently depend on predefined tools or early-stage tool generation, limiting their adaptability and scalability to complex scientific tasks. We introduce CASCADE, a self-evolving agentic framework representing an early instantiation of the transition from "LLM + tool use" to "LLM + skill acquisition". CASCADE enables agents to master complex external tools and codify knowledge through two meta-skills: continuous learning via web search, code extraction, and memory utilization; self-reflection via introspection, knowledge graph exploration, and others. We evaluate CASCADE on SciSkillBench, a benchmark of 116 materials science and chemistry research tasks. CASCADE achieves a 93.3% success rate using GPT-5, compared to 35.4% without evolution mechanisms. We further demonstrate real-world applications in computational analysis, autonomous laboratory experiments, and selective reproduction of published papers. Along with human-agent collaboration and memory consolidation, CASCADE accumulates executable skills that can be shared across agents and scientists, moving toward scalable AI-assisted scientific research.