Blog

RLinf-Co: Sim-Real Co-Training for VLA with RL

vlareinforcement-learningsim-to-real

RLinf-Co: Sim-Real Co-Training for VLA with RL

A hands-on guide to RLinf-Co — open-source framework combining RL and real data to train π₀/π₀.₅ VLAs far beyond pure imitation learning.

6/19/202610 min read

dream-chunkvlaaction-chunking

DREAM-Chunk for VLA Robots

A practical guide to DREAM-Chunk, reactive action chunking with a latent world model for robust VLA robot execution.

6/19/202614 min read

ENPiRE: Coding Agents That Teach Robots to Improve

nvidiagear-labcoding-agents

research

ENPiRE: Coding Agents That Teach Robots to Improve

NVIDIA GEAR Lab's ENPiRE uses coding agents (Claude Code, Codex, Kimi) to run real robot experiments autonomously, 99% pass rate on dexterous manipulation.

6/18/202610 min read

memoryvla++vlaworld-model

MemoryVLA++: Memory and World Models

A practical guide to MemoryVLA++: PCMB memory, latent imagination with a world model, training, inference, and long-horizon results.

6/17/202613 min read

hy-embodiedvlaumi

Run Hy-Embodied-0.5-VLA with UMI

A practical guide to installing, running, fine-tuning, and evaluating Hy-Embodied-0.5-VLA for bimanual robot manipulation.

6/17/202614 min read

Dexora: Open-Source VLA for Bimanual Dexterous Robots

vlabimanualdexterous-manipulation

manipulation

Dexora: Open-Source VLA for Bimanual Dexterous Robots

Dexora: the first open-source VLA for 36-DoF dual-arm, dual-hand manipulation, pairing an exoskeleton with Apple Vision Pro — 66.7% on dexterous tasks.

6/15/202610 min read

weaverworld-modelpi0.5

WEAVER: World Model for π0.5 VLA

A practical guide to WEAVER, a multi-view world model for evaluating, improving, and steering π0.5 VLA manipulation.

6/15/202611 min read

m3imicunitree-g1isaac-lab

M3imic: Multimodal WBC for G1

A beginner-friendly guide to M3imic: idea, architecture, Isaac Lab setup, training, inference, and Unitree G1 results.

6/15/202614 min read

Deep Dive

soma-xnvidiahumanoid

humanoid

SOMA-X v0.2: One Rig, Every Body

SOMA-X v0.2 unifies body models, pose conversion, LODs and pose inversion for physical AI, humanoid simulation and robotics.

6/13/202612 min read

riseworld-modelvla

RISE: Self-Improving Robot Policy

RISE uses a compositional world model so VLA policies can improve through imagined rollouts before real robot deployment.

6/13/202615 min read

riseworld-modelvla

RISE: Hands-on Training Pipeline Guide

Step-by-step walkthrough of the RISE training pipeline: environment setup, LeRobot data preparation, offline policy training, dynamics model, and online self-im

6/13/202612 min read

rovevlareinforcement-learning

ROVE: Human Intervention as RL Signal for VLA Humanoid

ROVE (XPENG Robotics) applies Optimistic Value Estimation to turn imperfect humanoid teleoperation data into RL signal for iterative VLA post-training.

6/22/202611 min read

vlareinforcement-learningsim-to-real

RLinf-Co: Sim-Real Co-Training for VLA with RL

A hands-on guide to RLinf-Co — open-source framework combining RL and real data to train π₀/π₀.₅ VLAs far beyond pure imitation learning.

6/19/202610 min read

dream-chunkvlaaction-chunking

DREAM-Chunk for VLA Robots

A practical guide to DREAM-Chunk, reactive action chunking with a latent world model for robust VLA robot execution.

6/19/202614 min read

nvidiagear-labcoding-agents

research

ENPiRE: Coding Agents That Teach Robots to Improve

NVIDIA GEAR Lab's ENPiRE uses coding agents (Claude Code, Codex, Kimi) to run real robot experiments autonomously, 99% pass rate on dexterous manipulation.

6/18/202610 min read

memoryvla++vlaworld-model

MemoryVLA++: Memory and World Models

A practical guide to MemoryVLA++: PCMB memory, latent imagination with a world model, training, inference, and long-horizon results.

6/17/202613 min read

hy-embodiedvlaumi

Run Hy-Embodied-0.5-VLA with UMI

A practical guide to installing, running, fine-tuning, and evaluating Hy-Embodied-0.5-VLA for bimanual robot manipulation.

6/17/202614 min read

vlabimanualdexterous-manipulation

manipulation

Dexora: Open-Source VLA for Bimanual Dexterous Robots

Dexora: the first open-source VLA for 36-DoF dual-arm, dual-hand manipulation, pairing an exoskeleton with Apple Vision Pro — 66.7% on dexterous tasks.

6/15/202610 min read

weaverworld-modelpi0.5

WEAVER: World Model for π0.5 VLA

A practical guide to WEAVER, a multi-view world model for evaluating, improving, and steering π0.5 VLA manipulation.

6/15/202611 min read

m3imicunitree-g1isaac-lab

M3imic: Multimodal WBC for G1

A beginner-friendly guide to M3imic: idea, architecture, Isaac Lab setup, training, inference, and Unitree G1 results.

6/15/202614 min read

Deep Dive

soma-xnvidiahumanoid

humanoid

SOMA-X v0.2: One Rig, Every Body

SOMA-X v0.2 unifies body models, pose conversion, LODs and pose inversion for physical AI, humanoid simulation and robotics.

6/13/202612 min read

riseworld-modelvla

RISE: Self-Improving Robot Policy

RISE uses a compositional world model so VLA policies can improve through imagined rollouts before real robot deployment.

6/13/202615 min read

riseworld-modelvla