RAVEN: real-time autoregressive video extrapolation
RAVEN is a training-time test framework that repacks self rollouts into interleaved clean …
Read moreLast Posts
MVP Engine: agentic training recipes for multimodal models
MVP Engine is a lightweight multimodal training framework that keeps orchestration small, moves …
Read moreLLaVA-OneVision-2: open long-video multimodal training
LLaVA-OneVision-2 extends fully open multimodal training toward long-video understanding, using …
Read moreOpen Roles
MLLM Research
Work on MLLM training, model architecture, and open-source training framework optimization for frontier multimodal systems.
MLLM Infra Engineering
Build the systems/frameworks that makes MLLM training, serving, and model engineering easy-to-use and efficient at scale.