---
title: "从开始就注定失败：通过受召回控制的探针级联早期中止LLM智能体回合"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.06503v1"
permalink: "https://feed.astralor.com/items/itm_1f1efec6d0651c86ffc4ef86"
published: "2026-07-07T17:03:55.000Z"
score: 78
channel: "tech"
tags: ["LLM Agent","Qwen-2.5-7B","内部表征探针","推理加速","早期中止","西北大学"]
---

# 从开始就注定失败：通过受召回控制的探针级联早期中止LLM智能体回合

## AI 点评

> 这篇工作巧妙地将失败预测从行为层面深入到内部表征，用轻量级探针实现早期中止，节省大量算力。对于部署多轮Agent服务的技术团队来说，是降低成本的利器。论文还给出了样本复杂度指引，让团队知道多少数据才够用，非常务实。

## 速览

多步LLM智能体经常陷入注定失败的执行轨迹，却消耗大量推理算力。本文提出一种基于内部表征的早期中止级联：每轮使用轻量探针读取隐藏激活，在首次交互即可预测失败，并设计受召回率控制的级联门控，确保成功回合以用户指定的全局召回率通过。在TextCraft上，该方法在90%召回目标下节省47.1%（Qwen-2.5-7B）和37.2%（Llama-3.2-3B）推理算力，远超单门策略和仅基于行为的方案。

## 关键点

- 利用LLM智能体内部隐藏激活进行早期失败预测，首次交互即可预见结局。
- 提出受召回控制的级联门控机制，保证成功回合以全局召回率通过。
- 在TextCraft上实现47.1%和37.2%的推理算力节省，且仅使用行为特征无额外增益。
- 分析并给出了保证高召回目标所需的样本复杂度边界。

## 全文

Large language model (LLM) agents solving multi-step tasks frequently commit to trajectories that are doomed to fail, yet continue to consume substantial inference compute before the failure becomes observable. We show that failure is predictable early from the agent's internal representations: lightweight per-round probes on hidden activations anticipate eventual episode failure as early as the first interaction round, where scorers reading only the agent's observable behavior are barely better than chance. We turn this signal into a practical abort cascade: one distribution-free calibrated gate per round, with per-round recall budgets jointly searched so that eventually-successful episodes survive all gates at a user-specified global rate; this episode-level guarantee is the one that matters in deployment, since false-abort risk accumulates across gates. Across two agent models on TextCraft, the cascade meets every recall target from 90% to 97% and, at the 90% target, saves 47.1% +/- 10.3% (Qwen-2.5-7B) and 37.2% +/- 8.8% (Llama-3.2-3B) of inference compute, 1.6--1.7x the best single-gate policy. An otherwise-identical cascade reading only behavior saves roughly half as much, and adding behavioral features to the probe yields no further gain: the hidden states capture what behavior reveals. Finally, we characterize the sample complexity of certifying high recall targets, telling practitioners which recall promises their data can, and provably cannot, back. The code will be released soon.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_1f1efec6d0651c86ffc4ef86