SAMBAR: Selective Anchoring via Method of Multipliers for Balanced Knowledge Acquisition and Retention in Vision-Language-Action Models

2026-09-26

AI Summary

This study addresses the catastrophic forgetting problem encountered by vision-language-action models during continual learning of new tasks without access to historical demonstration data. To this end, we propose SAMBAR, an algorithm that formulates continual learning as a constrained optimization problem. By leveraging the method of multipliers to dynamically adjust penalty coefficients and incorporating a selective parameter regularization strategy for model fine-tuning, SAMBAR effectively balances the acquisition of new knowledge with the retention of previously learned skills. Experimental evaluations on both the LIBERO simulation benchmark and real-world hardware demonstrate that existing replay-free baselines completely forget the initial task, whereas SAMBAR successfully preserves capabilities across all learned tasks, substantially improving the continual learning performance of vision-language-action models.

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.

Problem

• Vision-Language-Action models
• continual learning
• catastrophic forgetting
• replay-free

Innovation

• Continual Learning
• Method of Multipliers
• Selective Anchoring
• Catastrophic Forgetting
• Vision-Language-Action Models