Alibaba AI Framework Slashes Token Use by 99%
Researchers at Alibaba have developed SkillWeaver, a new AI framework designed to improve how AI agents select and combine tools for complex tasks. The approach significantly reduces computational resource consumption.
Alibaba researchers have unveiled SkillWeaver, an AI framework addressing the challenge of AI agents needing to efficiently select and orchestrate multiple tools for complex tasks. The framework constructs an execution graph for a given task, identifying the most suitable skills for each step.
The system incorporates Skill-Aware Decomposition (SAD), a technique that uses a feedback loop. This allows the AI agent to iteratively retrieve and vet relevant tool candidates. This compositional approach differentiates SkillWeaver from earlier methods that opt for tools in a single step.
SkillWeaver is applicable in scenarios where AI agents autonomously manage multi-tool ecosystems for operations like downloading datasets, transforming information, and generating visual reports. Experiments demonstrated that SkillWeaver reduced token consumption by over 99% compared to exposing agents to an entire tool library.
The research highlights that the granularity of task decomposition is the primary bottleneck for accurate tool retrieval. SkillWeaver breaks down complex requests into smaller sub-tasks, searches for appropriate tools, and builds a final execution plan, ensuring tool compatibility.