Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Towards Long-Lived Robots: Continual Learning VLA Models via Reinforcement Fine-Tuning

About

Pretrained on large-scale and diverse datasets, VLA models demonstrate strong generalization and adaptability as general-purpose robotic policies. However, Supervised Fine-Tuning (SFT), which serves as the primary mechanism for adapting VLAs to downstream domains, requires substantial amounts of task-specific data and is prone to catastrophic forgetting. To address these limitations, we propose LifeLong-RFT, a simple yet effective Reinforcement Fine-Tuning (RFT) strategy for VLA models independent of online environmental feedback and pre-trained reward models. By integrating chunking-level on-policy reinforcement learning with the proposed Multi-Dimensional Process Reward (MDPR) mechanism, LifeLong-RFT quantifies the heterogeneous contributions of intermediate action chunks across three dimensions to facilitate policy optimization. Specifically, (1) the Quantized Action Consistency Reward (QACR) ensures accurate action prediction within the discrete action space; (2) the Continuous Trajectory Alignment Reward (CTAR) aligns decoded continuous action chunks with reference trajectories to ensure precise control; (3) the Format Compliance Reward (FCR) guarantees the structural validity of outputs. Comprehensive experiments across SimplerEnv, LIBERO, and real-world tasks demonstrate that LifeLong-RFT exhibits strong performance in multi-task learning. Furthermore, for continual learning on the LIBERO benchmark, our method achieves a 22% gain in average success rate over SFT, while effectively adapting to new tasks using only 20% of the training data. Overall, our method provides a promising post-training paradigm for VLAs.

Yuan Liu, Haoran Li, Shuai Tian, Yuxing Qin, Yuhui Chen, Yupeng Zheng, Yongzhen Huang, Dongbin Zhao• 2026

Related benchmarks

TaskDatasetResultRank
Robot ManipulationSimplerEnv WidowX Robot tasks (test)
Success Rate (Spoon)84.3
79
Robot ManipulationSimplerEnv Google Robot tasks Visual Matching
Pick Coke Can Success Rate94
62
Multi-task LearningLIBERO
Object Score99.2
18
Continual LearningLIBERO Object
FWT96
8
Continual LearningLIBERO Goal
FWT92.4
8
Continual LearningLIBERO Spatial
FWT94
6
Continual LearningLIBERO Long
Forward Transfer (FWT)74.2
6
Continual LearningReal-world
FWT80
4
Hang Chinese KnotReal-world 1.0 (test)
Success Rate75
4
Multi-Task Learning (Overall)Real-world 1.0 (test)
Success Rate87.5
4
Showing 10 of 13 rows

Other info

Follow for update