VLA models exhibit catastrophic forgetting on a new real-world dataset of four sequential manipulation tasks, with experience replay implementation factors evaluated for mitigation.
Towards Long-Lived Robots: Continual Learning VLA Models via Reinforcement Fine-Tuning
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
abstract
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 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. The project page is available at <https://yuan-liu-lifelong-rft.github.io>.
citation-role summary
background 1
citation-polarity summary
years
2026 4verdicts
UNVERDICTED 4roles
background 1polarities
background 1representative citing papers
VLAs-as-Tools pairs a VLM planner with specialized VLA executors via a new interface and Tool-Aligned Post-Training to raise long-horizon robot success rates on LIBERO-Long and RoboTwin benchmarks.
ConSFT is a gradient-scaling fine-tuning objective for flow-matching VLAs that bounds parameter disruption via model-confidence weighting, yielding over 20% better capability retention than vanilla SFT on LIBERO and RoboTwin.
A closed-loop system couples LLM-based 3D scene generation with RL optimization and VR user interactions to produce adaptive, immersive environments, claiming SOTA results on the ALFRED benchmark.
citing papers explorer
-
Can VLA Models Learn from Real-World Data Continually without Forgetting?
VLA models exhibit catastrophic forgetting on a new real-world dataset of four sequential manipulation tasks, with experience replay implementation factors evaluated for mitigation.
-
Towards Long-horizon Embodied Agents with Tool-Aligned Vision-Language-Action Models
VLAs-as-Tools pairs a VLM planner with specialized VLA executors via a new interface and Tool-Aligned Post-Training to raise long-horizon robot success rates on LIBERO-Long and RoboTwin benchmarks.
-
Preserving Foundational Capabilities in Flow-Matching VLAs through Conservative SFT
ConSFT is a gradient-scaling fine-tuning objective for flow-matching VLAs that bounds parameter disruption via model-confidence weighting, yielding over 20% better capability retention than vanilla SFT on LIBERO and RoboTwin.
-
Closing the Loop: Unified 3D Scene Generation and Immersive Interaction via LLM-RL Coupling
A closed-loop system couples LLM-based 3D scene generation with RL optimization and VR user interactions to produce adaptive, immersive environments, claiming SOTA results on the ALFRED benchmark.