AgentProofBench 发布：面向 Web Agent 可复用验证器的基准
AgentProofBench: Benchmarking Reusable Verifiers for Web Agents
arXiv 上新出的 AgentProofBench，专门测 Web Agent 的验证器好不好用，用 DeepSeek 跑了近百个任务，做 Agent 评测的可以看看。
AgentProofBench 是一个评测 Web Agent 可复用验证规则的基准，覆盖 97 个任务、79 个任务族和 4 种 Web 服务。该基准将冻结的验收规则与从相同保存状态出发的执行过程配对测试。实验使用 DeepSeek Flash/Pro 两种模型，在 48 个任务族上，完整面板与决策回放测试分别保留 82/96 Flash 和 79/96 Pro 的源程序。在 22 个执行任务族中，面板无错误的 33 个程序分配里有 32 个正确交付，而有面板错误的 11 个里仅 1 个正确交付。
AgentProofBench: Benchmarking Reusable Verifiers for Web Agents
Reusable Web-agent verification requires a rule to follow changing requirements and support subsequent execution. We introduce AgentProofBench, which connects paired tests of a frozen acceptance rule to execution from identical saved openings. The resource spans 97 tasks, 79 families, and four Web services. Experiments with DeepSeek Flash/Pro establish two linked measurements. Across 48 families, complete-panel and decision-replay tests retain 82/96 Flash and 79/96 Pro source programs; paired reuse-rebuild BA intervals are [-5.47, 1.57] and [-10.16, 0.00] percentage points. Across 22 execution families, deployment-state panel correctness separates program assignments whose two saved openings are both delivered correctly: 32/33 assignments with error-free panels, versus 1/11 with panel errors. All reference-defined program-gate regressions and withheld deliveries occur in the latter group. Neutral continuation attains the highest correct delivery under the tested generic feedback, providing a concrete baseline for measuring the value added by a gate. Released instructions, states, programs, reference rules, and trajectories connect each test failure to its execution consequences. AgentProofBench makes reusable verification assessable at both the requirement and policy levels.