On December 19, 2024, at a press conference, ZhiYuan Research Institute and Tencent announced the launch of LongBench v2, a benchmark designed to evaluate the deep understanding and reasoning capabilities of large language models (LLMs) in real-world long text multi-task scenarios. The platform aims to advance long text models in understanding and reasoning, addressing the current challenges faced by large language models in practical applications.

On December 18, 2024, ZhiYuan Research Institute signed a strategic cooperation agreement with Tencent Group, aiming for in-depth collaboration in several areas, including large model research and development, exploration of cutting-edge artificial intelligence technologies, and open-source ecosystem construction. According to the cooperation agreement, ZhiYuan Research Institute and Tencent will leverage their respective strengths to promote the deep integration of large model technology and industrial scenarios, and explore technological solutions for large model training and inference optimization in diverse computing environments. Both parties are also committed to building an open and innovative hardware and software ecosystem to foster the development and application of technologies.

On December 16, ZhiYuan Robotics announced the official launch of commercial mass production of general-purpose robots, marking the company's transition from the startup phase to industrialization. Since its establishment in February 2022 in the Shanghai Lingang New Area, ZhiYuan Robotics has rapidly grown and released its first humanoid robot, 'Yanzheng A1'. This year, the company plans to achieve large-scale production of robots by 2024 and has built Shanghai's first humanoid robot mass production facility in Lingang Fengxian. Founder Peng Zhihui expressed on social media: 'In two years, we've forged a sword, from inspiration to blueprint, from prototype to mass production, ZhiYuan.'

The Beijing Academy of Artificial Intelligence (BAAI) recently announced the launch of an innovative 3D generation model, See3D, which can learn from large-scale unlabeled internet videos. This technological breakthrough marks an important step towards the concept of 'See Video, Get 3D'. The See3D model does not rely on traditional camera parameters but utilizes visual conditioning techniques to generate camera-direction controllable and geometrically consistent multi-view images using only visual cues from the video. This approach avoids the need for expensive 3D or camera calibration.

The Beijing Academy of Artificial Intelligence (BAAI) recently announced the launch of a brand-new all-in-one visual generation model, OmniGen, marking a significant breakthrough in the field of image generation. The OmniGen model is known for its unification, simplicity, and cross-task knowledge transfer capabilities, allowing it to handle multiple image generation tasks within a single framework, including text-to-image generation, image editing, theme-driven generation, and visual condition generation. Furthermore, OmniGen can also tackle classic computer vision tasks such as image denoising and edge detection.

The Beijing ZhiYuan Artificial Intelligence Research Institute, in collaboration with Shanghai Jiao Tong University, Renmin University of China, Peking University, and Beijing University of Posts and Telecommunications, has launched an ultra-long video understanding model named Video-XL. This model is an important demonstration of the core capabilities of multimodal large models and a key step towards General Artificial Intelligence (AGI). Compared to existing multimodal large models, Video-XL shows superior performance and efficiency in processing long videos exceeding 10 minutes.

Beijing Zhiyuan Artificial Intelligence Research Institute announces the launch of the native multimodal world model Emu3. This model is based on next-token prediction technology and does not rely on diffusion models or combinatorial methods to achieve understanding and generation across text, image, and video modalities. Emu3 surpasses existing well-known open-source models such as SDXL, LLaVA, and OpenSora in tasks like image generation, video generation, and visual language understanding, showcasing exceptional performance.

Beijing Baidu Network Technology Co., Ltd. and Beijing Zhiyuan Artificial Intelligence Research Institute announced today the formal signing of a strategic cooperation agreement, under which the two sides will engage in deep collaboration in large model and other fields, jointly constructing an AI industry-research collaborative ecosystem. Baidu has been comprehensively laying out its artificial intelligence strategy since 2010 and is one of the few companies globally that has fully integrated its AI efforts, from Kunlun chips, PaddlePaddle deep learning platform, Wenxin large model to applications.

Zhiyuan Research Institute has launched a dataset called Infinity-Instruct, which comprises millions of instruction tuning data aimed at enhancing the performance of language models in dialogues and other areas. Recently, Infinity-Instruct has completed a new round of iterations, including the Infinity-Instruct-7M foundational instruction dataset and the Infinity-Instruct-Gen dialogue instruction dataset.

Beijing ZhiYuan Artificial Intelligence Research Institute, in collaboration with Renmin University of China's Gaoqing School of Artificial Intelligence, has launched an innovative AI model framework called MemoRAG. This framework, based on long-term memory, aims to advance Retrieval-Augmented Generation (RAG) technology to handle more complex tasks beyond simple question answering.

On September 4, 2024, the Beijing ZhiYuan Artificial Intelligence Research Institute (BAAI) announced the launch of the world's first model combat evaluation service featuring text-to-video, called FlagEval Large Model Arena. This service is open to users and covers approximately 40 large models from both domestic and international sources, supporting customizable online or offline evaluations for four major tasks: language Q&A, multimodal image and text understanding, text-to-image, and text-to-video.

The Zhiyuan Research Institute recently released three new vector models that excel in vector retrieval tasks and have set new best standards across multiple evaluation benchmarks. The three models are: BGE-EN-ICL: an English vector model that enhances the model's semantic expression capabilities by introducing task-related query-document examples as few-shot examples.

Beijing Zhiyuan Artificial Intelligence Research Institute, in collaboration with China Telecom Artificial Intelligence Research Institute, has launched an upgraded version of the Tele-FLM series of large models, including the 52B instruction model FLM-2-52B-Instruct and the trillion-parameter model Tele-FLM-1T. The FLM-2-52B-Instruct focuses on enhancing Chinese conversational abilities through instruction fine-tuning, achieving 90% of GPT-4's capabilities, based on the Tele-FLM-52B base model, utilizing specific datasets and parameter optimization.

Kunlun Tech and the Beijing Academy of Artificial Intelligence, Nanyang Technological University in Singapore, Peking University, and other institutions have jointly released a general computer control framework named Cradle. This AI framework enables intelligent agents (AI Agents) to control keyboards and mice directly like humans without special training, interact with any open-source or closed-source software, and not rely on any internal API. Cradle is the first AI framework that can handle

ZhiYuan Research Institute has released a code generation training dataset called TACO, aimed at providing more challenging training data and evaluation benchmarks for code generation models. TACO has advantages in terms of data scale, quality, and evaluation schemes, including a larger training and testing set, diverse problem-solving answers, and fine-grained labels. Experimental results show that current popular code generation models show significant differences compared to GPT-4 in TACO evaluations, indicating that there is still room for improvement in this field. TACO is not just a challenging...

["ZhiYuan Research Institute has released the new generation multimodal foundation model Emu2, pushing the boundaries of multimodal contextual learning capabilities.", "Emu2 surpasses Flamingo-80B and IDEFICS-80B, demonstrating excellent performance in few-shot multimodal understanding tasks.", "Emu2 achieves optimal performance in multiple few-shot understanding, visual question answering, and image generation tasks.", "Emu2-Chat realizes accurate understanding of text-image instructions, while Emu2-Gen offers flexible, controllable, high-quality images."]

ZhiYuan Research Institute, in collaboration with TuoSi and ZhongKe WenGe, has jointly established the 'Chinese Internet Corpus' (CCI). This corpus has undergone strict screening and cleaning, with a data scale of 104GB, covering the period from 2001 to 2023. ZhiYuan Research Institute will continue to expand data sources and improve data processing workflows to provide more high-quality and reliable data resources. The institute has also opened up other high-quality Chinese datasets, such as WUDAO corpus, COIG, and MTP. This initiative aims to support the big data and artificial intelligence industries.