Act First, Reason Later: Accelerating On-Policy Distillation for Multi-Turn Agents via Reference-Conditioned Inverse Dynamics
arXiv preprint · arXiv:2609.36608
arXiv preprint · arXiv:2609.36608
Technical report · arXiv:2609.14973
ICML 2026 · PMLR 306:163739–163766
IEEE TETCI · 9(6):3773–3787
TMLR · Transactions on Machine Learning Research, 2024