Search Results for "post"

...It accepts multimodal inputs—such as language and images—and uses a diffusion transformer architecture built upon vision-language encoders, enabling adaptive robot behaviors across diverse environments. It is designed to be customizable via post-training with real or synthetic data. The vision-language model remains frozen during both pretraining and finetuning, preserving language understanding and improving generalization. Streamlined MLP connection between vision encoder and LLM with added layer normalization.

MBDyn_sim_suite is a collection of free pre&post-processing tools and simulation models for the open-source multi-body analysis software MBDyn forming a general purpose simulation environment for structural dynamics with an emphasis on wind turbines.