arXiv：cs.LG· Artin Tajdini, Manpreet Kaur, Peng Qi·· 4 小时前
CIPR：跨接口策略正则化提升 LLM 智能体鲁棒性
One Task, Many Interfaces: Simple Cross-Interface Consistency Training Improves LLM Agent Robustness
AI 导读
针对 LLM 智能体在语义不变接口扰动下性能下降的问题，研究者提出自监督目标 Cross-Interface Policy Regularization（CIPR），通过将动作名替换为无信息标识符、在同一训练样本的两种接口下惩罚动作分布的 KL 散度，使策略在不同接口间保持一致。
正文
Abstract:Post-training binds the competence of an LLM agent to the interface it was trained on. Semantics-preserving perturbations of that interface, such as renaming an action or changing the notation, leave the underlying task unchanged yet make the agent worse at it. Scoring each sample independently means interface randomization only broadens this dependence from a single interface to the training interface distribution, so out-of-distribution perturbations still degrade the agent. We propose Cross-Interface Policy Regularization (CIPR), a self-supervised objective that replaces action names with uninformative identifiers and presents each training sample under two interfaces in which equivalent actions are identified by construction. CIPR penalizes the KL divergence between the action distributions induced under the two interfaces, so the policy behaves consistently across them. We evaluate CIPR in both supervised and reinforcement-learning post-training, where it improves robustness. On agentic function calling, held-out interface perturbations reduce the accuracy of a post-trained agent by nearly 25%, and CIPR recovers 10.5% while leaving accuracy on the unchanged interface intact.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.20576 [cs.LG] |
| (or arXiv:2610.20576v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20576 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Artin Tajdini [view email]
[v1]
Wed, 14 Oct 2026 17:56:06 UTC (372 KB)
来源：arXiv：cs.LG · arxiv.org