AI 中文总结
该综述调研约150个机器人策略验证器，按可用性和可信度比较，发现两者此消彼长，提出验证器自身有效性的三种度量及九个可检验指标。
AI 中文摘要
机器人策略的验证器读取候选行为并返回其执行效果的评分，既用于评估视觉-语言-动作策略，也用于训练它们。验证器涵盖从成功检测器和奖励模型到运行时监控器、安全过滤器和时序逻辑规范等多种类型。我们综述了约150个验证器，并沿两个属性进行比较。可用性衡量的是验证结果花费多少成本、在部署过程中多早获得验证结果，以及可以多频繁地请求验证结果。随着验证结果变得更便宜、更早和更密集，可用性提高。可信度衡量的是高分在多大程度上反映了任务完成情况。当判断变得可被博弈和自利时，可信度下降。我们根据判断的提供者将验证器分组：人工验证器、基于规则和形式的验证器、学习和预训练的验证器，以及模型内在验证器。在这四类中，我们发现随着可用性的提高，可信度下降。无论判断由谁提供，都没有免费的检查器。接着，我们考察了什么验证了验证器本身，以及高分在多大程度上具有信息量。文献中出现了三种度量：与人工标签的一致性、其训练策略的性能，以及在奖励黑客行为下的表现。最后，我们提出了九个指标，使验证器的声明可被检验，并为尚待构建的验证器提供了坐标。
英文摘要
A verifier for robot policies reads a candidate behavior and returns a score for how well it did, used both to evaluate vision-language-action policies and to train them. Verifiers range from success detectors and reward models to runtime monitors, safety filters, and temporal-logic specifications. We survey roughly 150 verifiers and compare them along two properties. Availability is how much a verdict costs, how early in a rollout the verdict arrives, and how often a verdict can be asked for. Availability rises as verdicts get cheaper, earlier, and denser. Credibility is how much a high score tells us about the task. Credibility falls as the judgment becomes gameable and self-serving. We group the verifiers by who supplies the judgment: human verifiers, rule-based and formal verifiers, learned and pretrained verifiers, and model-intrinsic verifiers. Across the four families, we find that credibility falls as availability rises. Regardless of who supplies the judgment, there is no free checker. We then examine what validates a verifier itself, and how much a high score tells us. Three measures appear in the literature: agreement with human labels, the performance of the policy it trains, and behavior under reward hacking. We close with nine metrics that make a verifier claim checkable, and coordinates for the verifiers still to be built.
CommentsSurvey. 33 pages, 5 figures, 7 tables, 202 references. Covers reward models, success and failure detection, temporal-logic and formal verification, world-model evaluation, and reward hacking. Project page: https://github.com/ZJUSCL/Awesome-Robot-Verifier
