---
title: "AV-AIVAT：非完美信息游戏中认证随时有效停止的 74 倍更廉价 Agent 评估"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.06362v1"
permalink: "https://feed.astralor.com/items/itm_109def1169d75773db287d44"
published: "2026-08-06T17:57:11.000Z"
score: 76
channel: "ai"
tags: ["Agent 评估","AIVAT","AV-AIVAT","置信序列","非完美信息博弈"]
---

# AV-AIVAT：非完美信息游戏中认证随时有效停止的 74 倍更廉价 Agent 评估

## AI 点评

> 我最欣赏这篇论文的地方，是它没有停留在理论层面，而是用 7 万多手 HUNL 实证把“省样本”直接变成了“早决策”。对于那些每天要比较十几版 Agent 的团队，这套方法能显著压缩评估周期；对在线统计研究者来说，EB-CS 的结构化边界也提供了一个可借鉴的框架。

## 速览

评估两个 Agent 孰强孰弱需对局至证据足以区分实力与运气，每局都耗资金、推理或专家时间。AV-AIVAT 结合 AIVAT 方差削减与随时有效置信序列，在非完美信息游戏中实现认证早期停止。在 15 个 LLM Agent 配置、71,439 手 HUNL 对局中，AIVAT 中位数降方差 54 倍；95% 置信度、±1 大盲精度下，原始结果需中位数 74 倍手数才停止。EB-CS 提供有限样本精确认证，使评估可审计、可复核。

## 关键点

- AIVAT 在 HUNL 对局中中位数降方差 54 倍，但本身不提供停止规则。
- AV-AIVAT 将 AIVAT 与随时有效置信序列结合，在 95% 置信度下将原始评估所需手数降至约 1/74。
- EB-CS 负责有限样本精确认证，并为 Leduc hold'em 建立结构化 payoff 界。
- 评估可在证据充分时立即停止，且留下可审计、可复核的完整记录。

## 全文

Deciding which of two agents is stronger means playing games until skill outweighs luck, and every game costs money, model inference, or expert time. Since the number of games needed is unknown, fixed-budget evaluations either keep paying after the result is settled or stop before the agents can be told apart, while naive optional stopping with an ordinary confidence interval invalidates the stated level. We make such an evaluation stop as soon as its evidence suffices, with the guarantee intact. The Action-Informed Value Assessment Tool (AIVAT) reduces variance in imperfect-information games through conditional mean-zero corrections, by a median $54\times$ across 15 LLM agent configurations spanning 71,439 paired Heads-Up No-Limit Hold'em (HUNL) hands, but does not say when to stop. We combine AIVAT with continuously monitored Confidence Sequences (CSs) into anytime-valid AIVAT (AV-AIVAT), whose online value model learns only from past games so that no game scores its own correction. At the nominal 95\% level and a target precision of $\pm1$ Big Blind, raw outcomes need a median $74\times$ as many hands as AIVAT-corrected outcomes to stop under the Asymptotic CS (AsympCS). Exact finite-sample certification uses the Empirical-Bernstein CS (EB-CS), which needs an independently justified bound on corrected payoffs. We establish such a bound structurally for Leduc hold'em and characterize a width floor set by the CS's bet cap and that bound, which governs how much of a variance gain becomes earlier stopping; the descriptive HUNL EB-CS runs show a median $1.37\times$ stopping-time ratio. AV-AIVAT turns variance reduction into efficient, auditable early stopping while separating asymptotic screening from exact certification, so an evaluation can stop the moment its evidence suffices and hand a third party everything needed to recheck the verdict at that very stopping time.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_109def1169d75773db287d44