LiveFood is a dataset consisting of over 5100 gourmet videos, encompassing the four domains of ingredients, cooking, presentation, and consumption. All videos are meticulously annotated by professionals, and a strict double-checking system is employed to further ensure the quality of annotations. We have also proposed the Global Prototype Encoding (GPE) model to address the incremental learning problem, which achieves competitive performance compared to traditional techniques.

RoboGen is an automatic robot learning product based on generative simulation. It automatically generates diverse tasks, scenarios, and training supervision to achieve large-scale robot skill learning. RoboGen possesses the ability to autonomously propose, generate, and learn, enabling it to constantly generate skill demonstrations related to various tasks and environments.

Eureka is a human-level reward design algorithm implemented by encoding large language models. It leverages the zero-shot generation, code writing, and context improvement capabilities of state-of-the-art language models (such as GPT-4) to evolve and optimize reward code. The generated rewards can be used to acquire complex skills through reinforcement learning. Eureka-generated reward functions outperform human-expert-designed reward functions in 29 open-source reinforcement learning environments, including 10 diverse robot morphologies. Eureka is also capable of flexibly improving reward functions to enhance the quality and safety of generated rewards. By combining with curriculum learning, using Eureka reward functions, we first demonstrate a simulated Shadow Hand performing a pen-twirling trick, skillfully manipulating the pen at a fast speed within a circle.

OpenPipe/OpenPipe is a project that transforms expensive prompts into affordable fine-tuned models. It creates personalized models using pre-trained models and custom training data to achieve more efficient text generation. The advantage of OpenPipe/OpenPipe lies in its ability to provide users with high-quality text generation results while lowering training costs. The project offers flexible pricing strategies and various pricing options to meet the needs of different users. Key features of OpenPipe/OpenPipe include: training with expensive prompts, generating customized models, efficient text generation, and reduced training costs.

LiveFood is a dataset consisting of over 5100 gourmet videos, encompassing the four domains of ingredients, cooking, presentation, and consumption. All videos are meticulously annotated by professionals, and a strict double-checking system is employed to further ensure the quality of annotations. We have also proposed the Global Prototype Encoding (GPE) model to address the incremental learning problem, which achieves competitive performance compared to traditional techniques.

RoboGen is an automatic robot learning product based on generative simulation. It automatically generates diverse tasks, scenarios, and training supervision to achieve large-scale robot skill learning. RoboGen possesses the ability to autonomously propose, generate, and learn, enabling it to constantly generate skill demonstrations related to various tasks and environments.

Eureka is a human-level reward design algorithm implemented by encoding large language models. It leverages the zero-shot generation, code writing, and context improvement capabilities of state-of-the-art language models (such as GPT-4) to evolve and optimize reward code. The generated rewards can be used to acquire complex skills through reinforcement learning. Eureka-generated reward functions outperform human-expert-designed reward functions in 29 open-source reinforcement learning environments, including 10 diverse robot morphologies. Eureka is also capable of flexibly improving reward functions to enhance the quality and safety of generated rewards. By combining with curriculum learning, using Eureka reward functions, we first demonstrate a simulated Shadow Hand performing a pen-twirling trick, skillfully manipulating the pen at a fast speed within a circle.

OpenPipe/OpenPipe is a project that transforms expensive prompts into affordable fine-tuned models. It creates personalized models using pre-trained models and custom training data to achieve more efficient text generation. The advantage of OpenPipe/OpenPipe lies in its ability to provide users with high-quality text generation results while lowering training costs. The project offers flexible pricing strategies and various pricing options to meet the needs of different users. Key features of OpenPipe/OpenPipe include: training with expensive prompts, generating customized models, efficient text generation, and reduced training costs.