arXiv：cs.AI· Jiaxuan Dai, Tianyi Huang·· 5 小时前AI 评分46
TwinCheck：面向有状态工具智能体的证据锚定负孪生验证
TwinCheck: Evidence-Grounded Negative-Twin Verification for Stateful Tool Agents
AI 导读
TwinCheck 是一种推理时验证策略，仅在轨迹满足与局部失败假设绑定的证据条件时才考虑替换，并构造负孪生反事实方案，只有在孪生通过结构检查且成对验证器在两种候选顺序下都更偏好它时才替换智能体的提议。在 159 个多轮 BFCL V4 任务上，该策略将 GPT-5.6 Sol 的任务成功率从 45.3% 提升至 58.5%（95% 置信区间 [8.2, 18.8]），未观察到成功转失败的退化。
正文
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. 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. For paired evaluation, exact replay holds the agent's parsed responses and actions fixed until the first accepted replacement, separating intervention effects from resampling. 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.
| Comments: | Accepted at the NeurIPS 2026 Workshop Who Verifies the Agents? Toward Reliable Agent Development; Accepted at the NeurIPS 2026 Third Workshop on Agents in the Wild: Safety, Security, and Beyond |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.26911 [cs.AI] |
| (or arXiv:2609.26911v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.26911 arXiv-issued DOI via DataCite |
Submission history
From: Tianyi Huang [view email]
[v1]
Tue, 22 Sep 2026 18:09:40 UTC (1,647 KB)
[v2]
Tue, 6 Oct 2026 17:58:20 UTC (1,647 KB)
来源：arXiv：cs.AI · arxiv.org