Another good paper. Interesting finding on the benefits of the agent harness.
// Automata from agent traces // How much of your agent's behavior comes from the model, and how much from the harness you wrapped around it? New work collapses...
新研究将智能体轨迹压缩为紧凑有限状态机,在12个公开数据集上仅用7-43个状态,以0.997的适应度重放留出数据,毫秒级构建。FSM状态上下文在下一步预测上全面优于Agent Workflow Memory,失败预测AUROC达0.94。作者认为行为拓扑更多由部署框架而非底层LLM塑造。
新研究将智能体轨迹压缩为紧凑有限状态机,在12个公开数据集上仅用7-43个状态,以0.997的适应度重放留出数据,毫秒级构建。FSM状态上下文在下一步预测上全面优于Agent Workflow Memory,失败预测AUROC达0.94。作者认为行为拓扑更多由部署框架而非底层LLM塑造。
Another good paper. Interesting finding on the benefits of the agent harness.
来源:elvis· x.com