TwinCheck: Evidence-Grounded Negative-Twin Verification for Stateful Tool Agents

TwinCheck：面向状态工具智能体的基于证据的负孪生验证

Abstract: A single locally plausible tool call can derail an otherwise successful agent trajectory. Suspicion alone does not justify intervention, because the replacement itself can introduce the very failure verification is meant to prevent.

摘要： 单个看似合理的工具调用可能会破坏原本成功的智能体轨迹。仅凭怀疑不足以进行干预，因为替换操作本身可能会引入验证旨在防止的故障。

We introduce TwinCheck, an inference-time verification policy that considers replacement only when the trace satisfies an evidence condition tied to a trace-local failure hypothesis. It constructs a trace-grounded counterfactual alternative, a negative twin, and replaces the agent’s proposal only if the twin passes structural checks and the pairwise verifier prefers it in both candidate orders.

我们引入了 TwinCheck，这是一种推理时验证策略，仅当轨迹满足与轨迹局部故障假设相关的证据条件时，才会考虑进行替换。它构建了一个基于轨迹的反事实替代方案（即“负孪生”），并且仅在孪生方案通过结构检查且成对验证器在两种候选顺序下均更倾向于该方案时，才会替换智能体的原始提议。

For paired evaluation, exact replay holds the agent’s parsed responses and actions fixed until the first accepted replacement, separating intervention effects from resampling.

在配对评估中，精确重放（exact replay）会固定智能体解析出的响应和动作，直到出现第一个被采纳的替换，从而将干预效果与重采样区分开来。

In the primary analysis of 159 multi-turn BFCL V4 tasks with complete exact-replay pairs, the complete policy raises task success for GPT-5.6 Sol from 45.3% to 58.5% (95% task-bootstrap CI [8.2, 18.8]), with no observed success-to-failure regressions. Together, these findings recast execution-boundary repair as a constrained comparison, making the counterfactual action itself the object of verification.

在对 159 个具有完整精确重放对的多轮 BFCL V4 任务进行的初步分析中，该完整策略将 GPT-5.6 Sol 的任务成功率从 45.3% 提高到了 58.5%（95% 任务自助法置信区间 [8.2, 18.8]），且未观察到从成功到失败的倒退。总之，这些发现将执行边界修复重新定义为一种受约束的比较，使反事实动作本身成为了验证的对象。