VLA 線 · 查看同日 AI 報告 →查看同日 AI 报告 →

2026-02-26

VLA 研究日報 Pulsar

LIVE

— AI 線今日無資料 —— AI 线今日无资料 —

VLA 線VLA 线 · cs.RO · cs.AI · cs.LG

Path-Decoupled Hyperbolic Flow Matching for Few-Shot Adaptation 跨模态小样本适应，无明确VLA应用路径 CS.CV
Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning [NVIDIA] 提出高效VLA推理方法，通过可语言化的潜在规划实现 CS.RO
On Robustness of Vision-Language-Action Model against Multi-Modal Perturbations 研究VLA模型对多模态扰动的鲁棒性，但无具体创新点 CS.RO
NRSeg: Noise-Resilient Learning for BEV Semantic Segmentation via Driving World Models BEV语义分割，与VLA相关性较弱 CS.RO
Performance Asymmetry in Model-Based Reinforcement Learning 基于模型的强化学习性能不对称性研究，与VLA关联不大 CS.RO
SimToolReal: An Object-Centric Policy for Zero-Shot Dexterous Tool Manipulation [Stanford|Bohg] 提出零样本灵巧工具操作策略，具有实际应用潜力 CS.RO
Learning Humanoid End-Effector Control for Open-Vocabulary Visual Loco-Manipulation 提出开放词汇视觉loco-manipulation的人形末端执行器控制方法 CS.RO
PMG: Parameterized Motion Generator for Human-like Locomotion Control 参数化运动生成器，与VLA直接相关性较弱 CS.RO
TouchGuide: Inference-Time Steering of Visuomotor Policies via Touch Guidance 通过触觉引导在推理时调整视觉运动策略，解决细粒度操作问题 CS.RO
DeLTa: Demonstration and Language-Guided Novel Transparent Object Manipulation 提出演示和语言指导的透明物体操作方法，填补领域空白 CS.RO
PegasusFlow: Parallel Rolling-Denoising Score Sampling for Robot Diffusion Planner Flow Matching 机器人扩散规划流匹配，但实验设计较弱 CS.RO
Human-Exoskeleton Kinematic Calibration to Improve Hand Tracking for Dexterous Teleoperation 人-外骨骼运动学校准，与VLA直接相关性较弱 CS.RO

2026-02-26

VLA 研究日報VLA 研究日报

15 篇 11 篇共 26 篇

🔧 技術技术

Practical VLA [NVIDIA] 2026-02-26

Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning

提出高效VLA推理方法，通过可语言化的潜在规划实现

cs.RO 閱讀原文

Practical VLA [Stanford|Bohg] 2026-02-26

SimToolReal: An Object-Centric Policy for Zero-Shot Dexterous Tool Manipulation

提出零样本灵巧工具操作策略，具有实际应用潜力

cs.RO 閱讀原文

Practical VLA 2026-02-26

Learning Humanoid End-Effector Control for Open-Vocabulary Visual Loco-Manipulation

提出开放词汇视觉loco-manipulation的人形末端执行器控制方法

cs.RO 閱讀原文

Practical VLA 2026-02-26

TouchGuide: Inference-Time Steering of Visuomotor Policies via Touch Guidance

通过触觉引导在推理时调整视觉运动策略，解决细粒度操作问题

cs.RO 閱讀原文

Practical VLA 2026-02-26

DeLTa: Demonstration and Language-Guided Novel Transparent Object Manipulation

提出演示和语言指导的透明物体操作方法，填补领域空白

cs.RO 閱讀原文

Practical VLA 2026-02-26

HALO: A Unified Vision-Language-Action Model for Embodied Multimodal Chain-of-Thought Reasoning

提出统一的VLA模型，用于具身多模态链式思维推理

cs.RO 閱讀原文

Practical VLA 2026-02-26

Notes-to-Self: Scratchpad Augmented VLAs for Memory Dependent Manipulation Tasks

提出增强记忆依赖操作任务的VLA方法，具有实际应用潜力

cs.RO 閱讀原文

Practical VLA 2026-02-26

GeCo-SRT: Geometry-aware Continual Adaptation for Robotic Cross-Task Sim-to-Real Transfer

几何感知的持续适应方法，解决跨任务sim-to-real迁移问题

cs.RO 閱讀原文

Practical VLA 2026-02-26

IG-RFT: An Interaction-Guided RL Framework for VLA Models in Long-Horizon Robotic Manipulation

提出交互引导的RL框架，适用于长时序机器人操作

cs.RO 閱讀原文

Practical VLA 2026-02-26

UniLACT: Depth-Aware RGB Latent Action Learning for Vision-Language-Action Models

提出深度感知的RGB潜在动作学习方法，提升VLA模型性能

cs.RO 閱讀原文

Practical VLA 2026-02-26

Global Prior Meets Local Consistency: Dual-Memory Augmented Vision-Language-Action Model for Efficient Robotic Manipulation

提出双记忆增强的VLA模型，提高机器人操作效率

cs.RO 閱讀原文

📖 背景閱讀背景阅读

Background VLA 2026-02-26

Path-Decoupled Hyperbolic Flow Matching for Few-Shot Adaptation

跨模态小样本适应，无明确VLA应用路径

cs.CV 閱讀原文

Background VLA 2026-02-26

On Robustness of Vision-Language-Action Model against Multi-Modal Perturbations

研究VLA模型对多模态扰动的鲁棒性，但无具体创新点

cs.RO 閱讀原文

Background VLA 2026-02-26

NRSeg: Noise-Resilient Learning for BEV Semantic Segmentation via Driving World Models

BEV语义分割，与VLA相关性较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

Performance Asymmetry in Model-Based Reinforcement Learning

基于模型的强化学习性能不对称性研究，与VLA关联不大

cs.RO 閱讀原文

Background VLA 2026-02-26

PMG: Parameterized Motion Generator for Human-like Locomotion Control

参数化运动生成器，与VLA直接相关性较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

PegasusFlow: Parallel Rolling-Denoising Score Sampling for Robot Diffusion Planner Flow Matching

机器人扩散规划流匹配，但实验设计较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

Human-Exoskeleton Kinematic Calibration to Improve Hand Tracking for Dexterous Teleoperation

人-外骨骼运动学校准，与VLA直接相关性较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

SpikePingpong: Spike Vision-based Fast-Slow Pingpong Robot System

基于尖峰视觉的乒乓球机器人系统，与VLA直接相关性较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

Squint: Fast Visual Reinforcement Learning for Sim-to-Real Robotics

快速视觉强化学习，但实验设计较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

ActionReasoning: Robot Action Reasoning in 3D Space with LLM for Robotic Brick Stacking

使用LLM进行机器人积木堆叠，但缺乏深度分析

cs.RO 閱讀原文

Background VLA 2026-02-26

Cooperative-Competitive Team Play of Real-World Craft Robots

多智能体合作竞争游戏，与VLA直接相关性较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

Grasp to Act: Dexterous Grasping for Tool Use in Dynamic Settings

动态环境下的灵巧抓取，但实验设计较弱

cs.RO 閱讀原文

Background VLA 2026-02-26

Efficient and Explainable End-to-End Autonomous Driving via Masked Vision-Language-Action Diffusion

端到端自动驾驶，但与VLA直接相关性较弱

hf-papers 閱讀原文

Background VLA 2026-02-26

Recursive Belief Vision Language Model

递归信念VLA模型，但摘要无具体技术路线

hf-papers 閱讀原文

Background VLA 2026-02-26

BFA++: Hierarchical Best-Feature-Aware Token Prune for Multi-View Vision Language Action Model

多视图VLA模型的最佳特征感知令牌剪枝，但贡献一般

hf-papers 閱讀原文

首頁首页 / VLA 日報VLA 日报 / 2026-02-26