CRAFT: Counterfactual-to-Interactive Reinforcement Fine-Tuning for Driving Policies

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Keyu, Ye, Nanfei, Wang, Yida, Sun, Wenchao, Zhao, Danqi, Cheng, Hao, Zheng, Sifa
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915983679029248
author Chen, Keyu
Ye, Nanfei
Wang, Yida
Sun, Wenchao
Zhao, Danqi
Cheng, Hao
Zheng, Sifa
author_facet Chen, Keyu
Ye, Nanfei
Wang, Yida
Sun, Wenchao
Zhao, Danqi
Cheng, Hao
Zheng, Sifa
contents Open-loop imitation learning has advanced modern autonomous driving policy architectures, but closed-loop deployment remains vulnerable to policy-induced distribution shift. Existing post-training paradigms exhibit fundamental trade-offs: closed-loop RL fine-tuning provides grounded feedback from executed actions but is constrained by the sparsity of informative events, whereas counterfactual fine-tuning provides dense supervision over candidate futures but inherits bias from imperfect future estimates. We introduce Counterfactual-to-Interactive Reinforcement Fine-Tuning (CRAFT), an on-policy framework that formulates closed-loop post-training as proxy-residual optimization. CRAFT uses group-normalized counterfactual advantages as a dense proxy for real closed-loop advantages and aligns this proxy with the closed-loop world through grounded residual correction from interaction-critical events. To stabilize adaptation, CRAFT regularizes the online policy toward an EMA teacher via asymmetric KL self-distillation. Theoretically, CRAFT decomposes the real closed-loop policy gradient into proxy and residual terms under the same visited-state distribution, reducing residual variance with an aligned proxy while mitigating proxy bias through grounded residual approximation. Empirically, CRAFT achieves the strongest closed-loop gains on Bench2Drive across hierarchical planning, vision-language-action, and vocabulary-scoring architectures. Ablations, scaling behavior, stability analyses, and transfer results further validate the complementary roles of dense counterfactual proxy and grounded residual correction. Project page: https://currychen77.github.io/CRAFT.
format Preprint
id arxiv_https___arxiv_org_abs_2605_04470
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CRAFT: Counterfactual-to-Interactive Reinforcement Fine-Tuning for Driving Policies
Chen, Keyu
Ye, Nanfei
Wang, Yida
Sun, Wenchao
Zhao, Danqi
Cheng, Hao
Zheng, Sifa
Machine Learning
Robotics
Open-loop imitation learning has advanced modern autonomous driving policy architectures, but closed-loop deployment remains vulnerable to policy-induced distribution shift. Existing post-training paradigms exhibit fundamental trade-offs: closed-loop RL fine-tuning provides grounded feedback from executed actions but is constrained by the sparsity of informative events, whereas counterfactual fine-tuning provides dense supervision over candidate futures but inherits bias from imperfect future estimates. We introduce Counterfactual-to-Interactive Reinforcement Fine-Tuning (CRAFT), an on-policy framework that formulates closed-loop post-training as proxy-residual optimization. CRAFT uses group-normalized counterfactual advantages as a dense proxy for real closed-loop advantages and aligns this proxy with the closed-loop world through grounded residual correction from interaction-critical events. To stabilize adaptation, CRAFT regularizes the online policy toward an EMA teacher via asymmetric KL self-distillation. Theoretically, CRAFT decomposes the real closed-loop policy gradient into proxy and residual terms under the same visited-state distribution, reducing residual variance with an aligned proxy while mitigating proxy bias through grounded residual approximation. Empirically, CRAFT achieves the strongest closed-loop gains on Bench2Drive across hierarchical planning, vision-language-action, and vocabulary-scoring architectures. Ablations, scaling behavior, stability analyses, and transfer results further validate the complementary roles of dense counterfactual proxy and grounded residual correction. Project page: https://currychen77.github.io/CRAFT.
title CRAFT: Counterfactual-to-Interactive Reinforcement Fine-Tuning for Driving Policies
topic Machine Learning
Robotics
url https://arxiv.org/abs/2605.04470