NVIDIA fine-tunes Cosmos Predict 2.5 with LoRA/DoRA for robot video generation

NVIDIA 通过 LoRA/DoRA 微调 Cosmos Predict 2.5 实现机器人视频生成

NVIDIA published a Hugging Face post on fine-tuning Cosmos Predict 2.5 with LoRA and DoRA to generate robot first-person videos from text prompts; the post does not disclose dataset size, training cost, or evaluation results.

Why it matters: HKR-H/K/R pass: the robot POV video angle is clickable, and LoRA/DoRA on Cosmos Predict 2.5 is a concrete mechanism. Missing dataset scale and metrics keep it in the low featured band.