ALTK-Evolve：用学习记忆让非前沿模型登顶τ²-bench客服榜
原标题：ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model
AI 摘要
Hugging Face 博客介绍 ALTK-Evolve 方法：用 GPT-4.1 分析客服智能体在 τ²-bench 训练任务上的成功与失败轨迹，自动提炼行为规则并追加到系统提示中，无需更新模型权重或人工标注。在零售领域，该方法使 GPT-4.1 智能体的 Pass^4 从 57.5% 提升至 67.5%，超过榜单上所有零售条目；航空领域 Pass^4 从 35.0% 提升至 40.0%。这表明通过“学习记忆”式提示增强，非前沿模型也能在客服任务上取得领先表现。
正文节选
ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model ALTK-Evolve learns and applies a playbook using GPT-4.1, outperforming the retail entries on τ²-bench, a customer service agent benchmark—without updating model weights. Using GPT-4.1, an older, non-reasoning model, a basic customer-service agent reaches 67.5% Pass^4 on τ²-bench retail — meaning it solves the task on all four independent attempts for 67.5% of tasks. It scores above ev