One Task, Many Interfaces: Simple Cross-Interface Consistency Training Improves LLM Agent Robustness
[Uczenie maszynowe] Autorzy: Artin Tajdini, Manpreet Kaur, Peng Qi
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…
- Kategoria arxiv
- cs.LG
- Kategoria
- Uczenie maszynowe