Bai Xiang's team at Huazhong University of Science and Technology, in collaboration with South China University of Technology and other institutions, launched the next-generation OCR evaluation benchmark OCRBench v2, which evaluated 58 mainstream multimodal large models worldwide.

Research teams from the Hong Kong University of Science and Technology and Zhejiang University launched VA-MoE, a model that uses phased training and variable index embedding mechanisms to guide different expert modules to focus on specific types of meteorological variables.

With the breakthrough of large language models in conversational text generation and the maturity of high-fidelity speech synthesis technology, podcasts, a form that is highly dependent on creators, are also being influenced by AI.

"Ml-simplefold: A Lightweight Protein Folding Prediction AI Model" is now available on the HyperAI official website in the "Tutorial" section. Run it with one click to get a new experience in protein generation based on the flow matching folding model!

Science fiction films have depicted the myriad possibilities of artificial intelligence in various ways. Let's review these classic films and see which AI fantasies are slowly coming to life.

HyperAI's "Tutorials" section has launched several open-source OCR model tutorials. Experience various features online, including multilingual text parsing, structural transformation, and scene text recognition. Come and try it out!

HyperAI has carefully compiled some excellent books in the field of AI. These books, highly recommended by bigwigs such as Musk and Sam Altman, will reveal to you the current strategies and future ethics of artificial intelligence.

Professor Mingda Li's team at MIT and others proposed the SCIGEN geometric structure constraint integration method, which can be adapted to any pre-trained generative diffusion model to generate target materials with special limitations.

SpikingBrain-1.0, a large-scale brain-like spike model based on intrinsic complexity, is now available on HyperAI's official website in the "Tutorials" section. Come and experience my country's first large-scale brain-like linear foundation model!

HyperAI has compiled a collection of popular remote sensing object detection datasets, covering medical imaging, urban issues, indoor scenes, agricultural diseases, and more. Download them with one click, and experience them now!

The team of Professor Li Jinjin and Professor Huang Fuqiang from the Artificial Intelligence and Microstructure Laboratory of Shanghai Jiao Tong University has developed a new AI material design model CGformer, which successfully breaks through the limitations of traditional crystal graph neural networks.

September 22-26 AI Paper Recommendations

Teachers and students from East China Normal University have developed a real-time catalog of AI Agent papers called "Awesome AI Agents Live," which brings together recent research results and provides functions such as search, filtering, and category navigation.

NVIDIA's research team launched ReaSyn, a synthesizable molecular projection framework that integrates reasoning capabilities. It regards the synthesis path as the CoT reasoning path of LLM, opening up a new path to solving the real-world problems of molecular synthesis.

A research team from the Hong Kong University of Science and Technology proposed a fusion neural network framework for predicting multi-metal binding sites in protein sequences.

"Wan2.2-S2V-14B: Cinematic Audio-Driven Video Generation" is now available in the "Tutorials" section of HyperAI's official website (hyper.ai). Come and create your own digital human!

At the 2025 Alibaba Cloud Conference, the Tongyi team announced six new models and a new brand. The Qwen3-Max, Qwen3-VL, and Qwen3-Coder models were the most impressive.

"RFdiffusion2: Protein Design Tool" is now available in the "Tutorials" section of HyperAI's official website (hyper.ai). Run it and experience the latest protein generation tool with just one click!

HyperAI has compiled a series of valuable and widely used tutorials and datasets from September 15th to 19th, covering fields such as biomedicine, image generation, language translation, logical reasoning, and more.

A joint research team from the Allen Institute for Artificial Intelligence (Ai2) and other institutions in the United States has demonstrated for the first time that machine learning weather models can generate highly skilled global seasonal forecasts.

September 15-19 AI Paper Recommendations

Google DeepMind, in collaboration with New York University and other institutions, systematically discovered new unstable singularities in three different fluid equations for the first time based on a machine learning framework and a high-precision Gauss-Newton optimizer.

The research results of DeepSeek-R1 appeared on the cover of Nature. As the first mainstream large-scale model to pass peer review in an authoritative journal, it also disclosed the training cost in addition to the technology.

The Indian Institute of Technology Bombay, in collaboration with the University of Maryland, has developed a prediction model based on convolutional neural networks and transfer learning, enabling early forecasts of extreme rainfall events.

The Hunyuan-MT-7B Translation Model Demo is now available in the Tutorials section of HyperAI's official website (hyper.ai). Challenge yourself to champion translation performance!

HyperAI has compiled a series of valuable and widely used tutorials and datasets from September 8th to 12th, covering a variety of fields such as image generation, biology, and logical reasoning.

September 8-12 AI Paper Selection

A research team from the Chinese Academy of Sciences and other institutions jointly proposed a new twin clustering framework scSiameseClu for interpreting single-cell RNA-seq data.

In early September, Apple was revealed to be interested in acquiring French startup Mistral AI, and shortly thereafter, semiconductor giant ASML led its Series C financing with 1.3 billion euros.

A joint team from Tsinghua University and MIT has systematically proposed for the first time a smart city planning framework driven by a large language model, aiming to build AI into an "intelligent planning assistant" for humans.