One Task, Many Interfaces: Simple Cross-Interface Consistency Training Improves LLM Agent Robustness

arXiv:2610.20576v1 Announce Type: cross 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.

Related

• SupportCal: Label-Free Calibration of Post-Trained LLMs via Reference Support and Corroboration
• Beyond Task Completion: Training Capable and Safe Computer-Use Agents
• Can Compact Language Models Search Like Agents? Distillation-Guided Policy Optimization for Preserving Agentic RAG Capabilities
• Toward Plasticity-Preserving KL Regularization for Capability Retention in LLM Reinforcement Learning
• ValueFlow: Measuring the Propagation of Value Perturbations in Multi-Agent LLM Systems

Source: arXiv cs.CL | 2026-10-15