外观
论文arXiv 2610.2327112 天前
NestedVLA：为 Vision-Language-Action 模型学习整合与生成技能
原文链接：NestedVLA: Learning to Consolidate and Generate Skills for Vision-Language-Action Models（arXiv 2610.23271 · 发布日期：2026-10-16）
NestedVLA: Learning to Consolidate and Generate Skills for Vision-Language-Action Models
NestedVLA用任务条件超网络生成动作专家参数，整合异构技能，并在三个RL基准上验证效果。
核心要点
- 异构任务与环境的数据分散，多任务训练易干扰，模型合并可能引入参数冲突或需改架构。
- 方法先独立学习技能，再事后整合；嵌套超网络按目标任务生成可与冻结基础VLA组合的动作专家参数。
- 在MetaWorld、ManiSkill、CALVIN上，平均较最强模型合并基线高41个百分点，并在ManiSkill接近逐任务微调。
- 任务特定压缩VLM残差被单独保留，代码已在GitHub公开。
验证范围与局限：验证范围为MetaWorld、ManiSkill、CALVIN三个RL基准，摘要未提及真机实验。
英文摘要 (Abstract)
Vision--Language--Action (VLA) models promise generalist robotic agents, yet consolidating skills learned across heterogeneous tasks and environments remains difficult: training data is fragmented across platforms, joint multi-task training suffers from interference, and model merging tends to introduce parameter conflicts or require architectural changes. We instead advocate a paradigm in which skills are first learned independently and then consolidated post hoc. To this end, we propose NestedVLA, which replaces parameter aggregation with parameter generation: a nested hypernetwork, conditioned on the target task, synthesizes task-adaptive action-expert parameters that compose with a frozen base VLA. Task-specific compressed VLM residuals are retained separately. This formulation enables flexible task-conditioned skill reuse while mitigating interference. Across three RL benchmarks (MetaWorld, ManiSkill, CALVIN), NestedVLA outperforms the strongest model-merging baseline by 41 pp on average, matches per-task fine-tuning within 1 pp on ManiSkill, and substantially closes the gap on CALVIN, pointing toward a storage-efficient route to skill consolidation. Code is available at https://github.com/WanyuGroup/NeurIPS2026-NestedVLA.
作者：Du Guodong, Wanyu Lin
来源
本页事实对应的公开页面，以原文为准。