跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 19 天前AI 評分31

DRAP:面向 Physical AI 持續學習的動態風險自適應可塑性方法

DRAP: Dynamics-Risk Adaptive Plasticity for Continual Learning in Physical AI

AI 導讀

DRAP 提出一種免重播持續學習方法,按軌跡放大風險調節模型組件的可塑性,以提升 Physical AI 對舊有行為的保留能力,同時學習新任務。方法在固定的 OpenVLA 子網絡上評估,並於 LIBERO Spatial、Object 和 Goal 測試套件進行序列適應;相比標準 AdamW 微調和 EWC,DRAP 改善了舊行為保留,並維持新測試套件的學習能力。

正文

Published on Nov 3

Authors:

Abstract

DRAP improves replay-free continual learning for physical AI by adaptively reducing plasticity of critical model components based on trajectory-amplification risk, enhancing retention of past behaviors without sacrificing new task learning.

Continual learning in physical AI differs from conventional continual learning because small per-step policy errors can accumulate over a trajectory. Each deviation may move the agent outside the training distribution, increase subsequent prediction errors, and ultimately cause closed-loop task failure. This work introduces Dynamics-Risk Adaptive Plasticity (DRAP), a replay-free method that estimates the historical importance of adaptable model tensors from sample-level action gradients weighted by local trajectory-amplification risk. A ridge-regularized local linear model is fitted to expert state-action transitions, and the largest singular value of its state-transition matrix provides a policy-independent, Lyapunov-inspired estimate of one-step perturbation amplification. After each task, risk-weighted squared gradients are accumulated in a tensor-level importance memory. During subsequent tasks, DRAP uses normalized cumulative importance to scale the effective AdamW learning rate of each tensor, immediately reducing the plasticity of components that supported previous behavior at dynamically sensitive trajectory points. We instantiate DRAP on a fixed KappaTune-selected subnetwork of OpenVLA and evaluate replay-free sequential adaptation across the LIBERO Spatial, Object, and Goal suites. Compared with standard AdamW fine-tuning and EWC, the results show substantially improved retention of previously learned behaviors while preserving strong new-suite acquisition.

View arXiv page View PDF GitHub 0 Add to collection

Get this paper in your agent:

hf papers read 2611.05029

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2611.05029 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.05029 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2611.05029 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co