外观
论文arXiv 2610.1195220 天前
NegaAlign框架以11.6M参数将VLA否定指令成功率从2.6%提至88%
本文由 AI 根据公开内容整理，可能有误，请以原文为准：Tell Robot What Not to Do: A Negation Understanding Perspective（arXiv 2610.11952）
核心内容
针对视觉-语言-动作模型（VLA）在执行自然语言指令时难以处理“不要做某事”等否定约束的问题，研究提出NegaAlign参数高效微调框架。该方法无需重训整个模型，仅通过图像-语言监督扩展预训练VLA能力。实验显示，在NegaBench仿真基准及真实世界任务中，搭载NegaAlign的π0.5模型否定指令执行成功率分别从2.60%和12.4%大幅提升至88.45%和88.8%，同时在GR00T、π0等主流模型上验证了通用性。
技术上，NegaAlign在视觉-语言骨干网络特定层插入Negation Transformation Layers重塑中间指令表征，并设计教师引导对齐机制，将满足否定约束的视觉Token与动作接地进行对齐。训练阶段冻结所有预训练参数（含动作生成器），仅更新插入层，可训练参数量仅为11.6M。配套发布的NegaBench涵盖5大领域10个场景，为评估操作任务中的否定约束遵循提供了标准化测试环境，解决了现有数据集缺乏此类负样本的难题。
关键看点
- 参数高效微调：仅新增11.6M可训练参数，冻结预训练VLA及动作生成器，通过插入变换层实现否定语义注入
- 性能跃升显著：π0.5模型在NegaBench仿真中否定指令成功率从2.60%提升至88.45%，真机实测达88.8%
- 教师对齐机制：利用教师模型引导视觉Token对齐，将否定约束转化为动作相关接地信号，避免错误执行
- 专用评测基准：发布NegaBench覆盖5域10场景，填补了具身智能否定指令遵循能力的系统化评估空白
工程与产业观察
该工作证明了在不破坏预训练VLA泛化能力的前提下，可通过极低成本适配解决安全关键型否定约束问题。对于机器人落地而言，这意味着无需海量负样本重训即可快速补齐“禁止操作”短板，显著降低人机交互中的安全风险与数据标注成本，为具身大模型的合规部署提供了轻量化技术路径。
英文摘要 (Abstract)
Instruction following enables robots to perform diverse tasks specified in natural language, making it a fundamental capability for human-robot interaction. Beyond communicating desired outcomes, users also need to specify constraints on what not to do. We investigate how to enable vision-language-action models (VLAs) to follow negated instructions, where robots must accomplish task goals while respecting explicit exclusions. To this end, we propose NegaAlign, a parameter-efficient, plug-and-play framework that extends pretrained VLAs to follow negated instructions through image-language supervision alone. Specifically, we introduce Negation Transformation Layers into selected layers of the vision-language backbone to reshape intermediate instruction representations. Meanwhile, a teacher-guided alignment mechanism is designed to align instruction-relevant visual tokens, transferring action-relevant grounding from instructions that satisfy the negated constraint. The training phase uses supervision constructed from existing demonstrations and updates only the inserted layers, keeping all pretrained parameters frozen, including the action generator. We further introduce NegaBench, a simulation benchmark spanning 10 scenarios across five domains for systematically evaluating manipulation under negated constraints. Experiments across GR00T, $π_0$, and $π_{0.5}$ demonstrate consistent improvements in negated instruction following. With 11.6M trainable parameters, NegaAlign increases the negated-instruction success rate of $π_{0.5}$ from 2.60% to 88.45% on NegaBench and from 12.4% to 88.8% on real-world tasks, while retaining performance on affirmative instructions.
作者：Fazeng Li, Gan Sun, Hao Cheng, Weihong Ren, Yang Cong
来源
本页事实对应的公开页面，以原文为准。