Technology ❯ Robotics ❯ Machine Learning

Vision-Language Models

Robotic Manipulation Benchmarking Gemini Robotics 1.5 Corruption Restoration Geometric Learning Vision-Language-Action Models 3D Point Cloud Processing Action Prediction Semantic Assembly Representation Zero-Shot Learning Spatial Reasoning Training and Scaling Primitive-Centric Framework DySL-VLA Framework Tactile Feedback Action Generation Flow Matching Models Geometric Integration Evaluation Methods Action Control In-Context Imitation Learning Tactile Feedback Systems

Three Papers Offer Data, Inference and Safety Fixes for Fragile VLA Robot Models

The proposals aim to cut shortcut learning, stop open-loop action failures, and predict collisions before they occur.

DeepMind Unveils Gemini Robotics‑ER 1.5 and Gemini Robotics 1.5 to Let Robots Plan Tasks and Use the Web