24 Latest AI Open Source Projects of the Week(2024.12.30-2025.1.5)
Explore 24 innovative AI open-source projects (2024.12.30–2025.1.5) covering LLMs, robotics, finance, medical AI, and creative tools. Discover groundbreaking tech!
I’m sharing some interesting AI open-source models and frameworks for this week (2024.12.30-2025.1.5).
There are a total of 24 AI open-source projects.
Project: SEMIKONG
SEMIKONG is an open-source large language model (LLM) specifically designed for the semiconductor field.
This project aims to address unique challenges in the semiconductor industry, such as the physical and chemical issues of semiconductor devices and processes, by integrating domain-specific knowledge into the model.
The SEMIKONG series models are bilingual language models trained on a 3T multilingual corpus, demonstrating strong capabilities in language understanding, commonsense reasoning, and reading comprehension.
https://github.com/aitomatic/semikong
Project: YuLan-Mini
YuLan-Mini is a lightweight language model with 2.4B parameters. Despite being pre-trained on only 1.08 trillion tokens, its performance rivals industry-leading models trained on larger datasets. The model excels particularly in mathematics and code-related tasks. To enhance reproducibility, the project will open-source the related pre-training resources.
https://github.com/RUC-GSAI/YuLan-Mini
Project: OD-FinLLM
OD-FinLLM is a fine-tuned model derived from the LLaMA series, focused on enhancing Chinese financial knowledge.
The model achieves outstanding performance in financial queries by fine-tuning LLaMA using a specially crafted instruction dataset created from publicly available Chinese financial Q&A data and additional web-scraped financial information.
https://github.com/Lattice-zjj/On-Device-FinLLM
Project: AgiBot-World
AgiBot-World is the world’s first large-scale, high-quality robotics manipulation benchmark, offering over one million trajectory data points from 100 robots across 100+ real-world scenarios, covering five target domains. The project leverages cutting-edge hardware, including vision-tactile sensors, 6-DOF dexterous hands, and mobile dual-arm robots, supporting contact-rich manipulation, long-term planning, and multi-robot collaboration tasks.
https://github.com/OpenDriveLab/agibot-world
Project: VideoVAEPlus
VideoVAEPlus is an advanced video variational autoencoder (VAE) designed for high-fidelity video reconstruction. The project improves reconstruction quality through cross-modal and joint video-image training, setting a new benchmark for video reconstruction tasks.
https://github.com/VideoVerses/VideoVAEPlus
Project: RID
The Real-time Identity Defenses (RID) project aims to provide real-time identity defenses against malicious personalization in diffusion models.
Built on the Diffusers library, this project offers a method to protect images from malicious personalization attacks.
Users can apply the defense mechanism with simple command-line operations to ensure image security.
https://github.com/Guohanzhong/RID
Project: HuatuoGPT-o1
HuatuoGPT-o1 is a medical large language model designed for advanced medical reasoning. It identifies errors, explores alternative strategies, and optimizes answers.
By leveraging verifiable medical queries and a dedicated medical validator, HuatuoGPT-o1 fine-tunes large language models through guided searches of complex reasoning paths and further enhances reasoning capabilities using reinforcement learning with reward models (PPO).