SMBAR is a continual learning algorithm that prevents catastrophic forgetting during VLA finetuning without requiring access to the demonstrations of any previously learned task, and is cast as a constrained optimization problem and solved with the method of multipliers.
Abstract
Vision-Language-Action (VLA) models leverage large-scale pretraining to ultimately achieve generalist manipulation. Deployed VLA policies must support continual learning to acquire new tasks over time. Teaching a VLA a new task generally requires finetuning it on demonstrations of that task. However, naively finetuning on downstream tasks causes the policy to forget earlier tasks and degrades generalist capabilities. This failure is known as catastrophic forgetting. Most continual learning methods counter it by replaying data from earlier tasks. However, the old task demonstrations are not always readily available. In this paper, we introduce SAMBAR, a continual learning algorithm that prevents catastrophic forgetting during VLA finetuning without requiring access to the demonstrations of any previously learned task. We propose to cast continual learning as a constrained optimization problem and solve it with the method of multipliers. In our approach, the method of multipliers drives the policy to learn the new task without the model parameters drifting far away from their previous values. In contrast to a standard regularization penalty, the method of multipliers raises the penalty as the constraint violation accumulates by using a dual variable. We also selectively anchor the parameters critical to previous tasks to preserve past knowledge, leaving other parameters free for new task acquisition. The combination of dual variable and selective anchoring, therefore, balances knowledge acquisition with knowledge retention. We evaluate our method, SAMBAR, on the LIBERO simulation benchmark and on hardware. When sequentially finetuning on a VLA, every replay-free baseline we compare against completely forgets the first task it learned, whereas SAMBAR retains every task it has learned.
Perturbot and GroundingFscore provide a training-and-evaluation framework for disentangling VLA decisions from shortcut priors while preserving responsiveness to task-relevant evidence.
Ming-Yu Liu, Chong-Hao Sima, Tianjian Feng et al.· 1 citation
ReDraft (Reference-Driven Revision and Fine-Tuning) shows that revising the model's own rollout rather than directly imitating an expert trajectory can reconcile cold-start acquisition with prior-capability retention.
FAN is introduced, which achieves the highest performance and demonstrates consistent robustness, providing insightful guidance for building stable action representations in achieving effective lifelong VLA adaptation.
Yi-Jun Hong, Jia-Run Zhu, Xiao-Quan Sun et al.· 0 citations
This paper proposes a self-supervised method that generates online interaction rollouts from the zero-shot VLA as additional training data for finetuning and demonstrates the success of this approach across test sets probing generalization on a real ALOHA robot and a new simulation benchmark in RoboTwin.
Prachi Garg, Steve Xing, Prahit Yaugand et al.· 0 citations
This paper introduces RouteRLT, a modular framework that coordinates a generalist VLA, used as the default controller, with designated precision-critical RL specialists, and trains a phase-aware coordination mechanism that handles handoffs between the generalist and the specialists.
Chong-Yu Zhu, Hyegang Kim, Jaden Hinds et al.· 0 citations
CorrectVLA is presented, a framework that translates task-level natural language corrections into additive action magnitude adjustments without modifying policy weights, and succeeds when policies possess strategic correctness and fails when fundamental comprehension is absent, establishing a practical operational boun...
O. Kwon, Pablo Ortega-Kral, A. Bucker et al.· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 9, 2026