让大模型代理学会预算感知，提前止损节省资源。
BAGEN: Are LLM Agents Budget-Aware?
- 将预算分为内部计算和外部动作两类，以区间预测方式动态监控
- 顶尖代理仍过度乐观，失败时继续耗资源，早停可省28%-64%令牌
- 预算预警信号可训练，适合需要成本控制的智能体应用
随着智能体资源消耗增加，当前成本多在执行后才被测量。预算感知代理（BAGEN）应将预算视为主动控制信号而非被动度量。我们首次系统定义预算估计算为内部预算（来自代理计算）和外部预算（来自代理行为）。进一步将预算感知形式化为渐进区间估计：每一步计划中，代理需预测剩余预算的上下界，并在完成可能性低时发出警告。通过回放式评估协议，在四个环境和五个前沿代理上发现一致失败模式：（1）强代理不一定具备强预算感知能力，相关性仅为 r=0.35。（2）前沿模型普遍过于乐观，对难以成功任务持续投入，而非早期预警。（3）预算感知信号具有可操作性和可训练性；早期停止在失败轨迹中节省28%-64%令牌，且SFT+RL能增强早停与预警行为。（4）精确区间校准仍具挑战，即使经过SFT+RL，区间覆盖率最高仅达47%。
原文摘要 · Abstract (English)
While agents are increasingly spending more resources, today agent cost is mostly measured only after execution. A Budget-Aware Agent (BAGEN) should treat budget as an active control signal, rather than a passive cost metric. We first systematically define budget estimation as internal budgets (from agent computation) and external budgets (from agent actions). We then formalize budget-awareness as progressive interval estimation: at each step of a plan, an agent should predict an upper and lower bound on remaining budget, and alert when completion is unlikely. Scoring with a rollout-replay protocol, we find consistent failure patterns on four environments and five frontier agents: (1) strong agents do not necessarily have strong budget-awareness, with correlation r=0.35. (2) frontier models are consistently over-optimistic, continue spending on tasks that are unlikely to succeed, instead of alerting the user early. (3) budget-aware signal is actionable and trainable. Early stop saves 28-64% tokens on failed trajectories, and SFT+RL strengthens early stop and alert behavior. (4) precise interval calibration remains challenging, with interval coverage capping at 47% after SFT+RL. Project page: https://ragen-ai.github.io/bagen/
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。