CASCADE: Cumulative Agentic Skill Creation through Autonomous Development and Evolution
This addresses the problem of limited adaptability and scalability of AI agents in complex scientific research, representing an incremental step toward more autonomous AI-assisted science.
The paper tackles the limitation of LLM agents' dependence on predefined tools by introducing CASCADE, a self-evolving framework that enables agents to acquire and codify skills through continuous learning and self-reflection, achieving a 93.3% success rate on a benchmark of 116 materials science and chemistry tasks compared to 35.4% without evolution mechanisms.
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.