Buildingthe way to multimodal intelligence.

Multimodality is good.

We believe it is a necessary condition for intelligence. So we build models that see, hear, and act in the world, and learn from feedback on their actions.

Explore Research

About the Lab

LMMS-LAB // BREACH ACTIVENEURAL WEIGHT EXTRACTION

1/7LIVE

thinking:

Inside the labModels · Evaluation · Open source

Featured Research

NOV 3, 2026

NEO-VLA:
Towards Native Vision-Language-Action Policies

NEO-VLA feeds pixels straight into a native vision-language backbone. Control draws on its shallow layers, and native pretraining drives the gains.

Latest Publications

LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence

models

JAN 2026

OneVision Encoder: Codec-Aligned Sparsity as a Foundational Principle for Multimodal Intelligence

models

SEP 2025

LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training

models

NOV 2025

LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling

models

Buildingthe way to multimodal intelligence.

NEO-VLA:Towards Native Vision-Language-Action Policies

LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence

OneVision Encoder: Codec-Aligned Sparsity as a Foundational Principle for Multimodal Intelligence

LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training

LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling

NEO-VLA:
Towards Native Vision-Language-Action Policies