Recently, Alibaba DAMO Academy (Hangzhou) Technology Co., Ltd.'s model open-source community, ModelScope, announced a significant breakthrough: the release of the InternVL 2.5 model. This open-source multimodal large language model has achieved exceptional performance, becoming the first open-source model to exceed 70% accuracy on the Multimodal Understanding Benchmark (MMMU), rivaling commercial models like GPT-4o and Claude-3.5-Sonnet.

In the arena of artificial intelligence, China is steadily advancing with a posture of independent innovation. On November 26, the world's first multimodal large model with 100 billion parameters, 'Zidong Taichu' 3.0, was officially unveiled in Wuhan, adding a significant chapter to domestic AI technology. This is not just a technological iteration, but an important milestone in China's AI ecosystem. Jointly developed by the Institute of Automation, Chinese Academy of Sciences, Huawei, and the Wuhan East Lake High-Tech Zone, 'Zidong Taichu' has successfully built the world's first pre-trained model covering image, text, and audio modalities.

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.

At a recent press conference, Alibaba International's AI team unveiled its latest development, the multimodal large model Ovis. This innovative AI technology undoubtedly brings new opportunities across various industries. Ovis boasts powerful image understanding and data processing capabilities that are refreshing. Ovis's multimodal ability is exceptionally strong, allowing it to handle various data types such as text and images, showcasing excellent comprehensive strength. Compared to traditional large language models, Ovis not only understands text but also conducts in-depth analysis of non-text information such as images.

On August 23, 2024, the well-known Chinese artificial intelligence company Yunzhisheng announced the launch of its latest research and development achievement – the Shanhai Multimodal Large Model in Beijing. The Shanhai Multimodal Large Model is part of Yunzhisheng's Atlas AI infrastructure, capable of receiving and processing inputs from various modalities, including text, audio, and images, and generating any combination of text, audio, and image outputs in real-time. This capability enables the Shanhai model to not only conduct efficient voice interactions but also provide a conversational experience that is close to natural human dialogue.

Recently, Moore Threads and Shiyin Intelligence announced a strategic partnership to jointly promote the application of domestically produced full-function GPUs in industry large model solutions. The two parties have completed the adaptation work of the Moore Threads Kuaguo Computing Cluster with Shiyin Intelligence's 'A Leaf Light Boat' multimodal large model, integrating their respective advantageous resources in the field of artificial intelligence, and are committed to providing a more intelligent and efficient service experience.

Recently, significant progress has been made in the research and application of multimodal large models. Foreign companies such as OpenAI, Google, and Microsoft have introduced a series of advanced models, and domestic institutions such as Zhipu AI and Jietu Xingchen have also made breakthroughs in this field. These models typically rely on visual encoders to extract visual features and integrate them with large language models, but they face issues such as visual inductive bias due to training

The "LingDong" brand has been meticulously crafted by NetEase Vuex, relying on its independently developed industrial-grade large models and AOP (Artificial Intelligence Operation Perception) technology concept. Two flagship products under the brand, the mining robot and the loading robot, have been put into use in over 50 provincial key projects, serving diverse environments such as mines, ports, mixing stations, and schools. In addition to the robot brand launch, NetEase Vuex also showcased it

OpenAI and Microsoft are being sued by news agencies for copyright infringement. Multimodal large models are becoming the mainstream trend in the field of large models. OpenAI is reportedly actively negotiating with publishers on AI policies while high-quality dataset copyright issues are receiving attention. CITIC Publishing collaborates with large model companies for language training, and Visual China holds a core advantage in the era of AIGC.

Kuaishou has launched a new AI strategy, developing large models and multimodal large models exceeding a scale of 100 billion. Kuaishou has introduced a text-to-image large model called 'Ke Tu', which can generate detailed and diverse styles of artwork. Kuaishou is accelerating the application of large models in various scenarios, providing more multimodal AI-generated content services. Kuaishou's products possess multiple AI-generated creative capabilities, assisting creators in easily producing content in different styles. Kuaishou is committed to maintaining its leading position in the domestic market to respond to potential upcoming technological breakthroughs.

SimoTech has launched the world's first industrial multimodal large model, IndustryGPT V1.0. This large model can accurately answer questions related to the manufacturing industry, identify industrial defects, and provide decision-making support. A strategic cooperation agreement has been signed with multinational companies to jointly promote the intelligent era of global industrial manufacturing. The release of this new model is of significant importance in the fields of intelligent manufacturing and digital innovation. The launch of the new model will bring technological innovation and development to the industrial manufacturing sector.

The Mianbi Intelligent multimodal large model application 'Mianbi Luca' is officially opened to public services. 'Mianbi Luca' is a multimodal intelligent dialogue assistant developed by Mianbi Intelligent based on its self-researched hundred-billion-parameter base model CPM. Luca possesses Chinese and English conversation capabilities, coding, knowledge, logic, and image understanding abilities. Mianbi Intelligent will further integrate AI Agent technology with Luca to leverage the potential of large model applications. Mianbi Intelligent collaborates with Tsinghua University and the OpenBMB open-source community to promote large...

The Ma Yi team proposed the EMT framework to evaluate the catastrophic forgetting of fine-tuned Multimodal Large Models (MLLMs). Experiments showed that while fine-tuning MLLMs improves performance on the fine-tuning dataset, it also leads to a decline in performance on other datasets. During the fine-tuning process, MLLMs generate hallucinated text related to the fine-tuning dataset, ignoring the original problem. This research provides a framework and benchmarks for future work, but the design and training techniques of the model still need further optimization. The Ma Yi team systematically assessed the issue of catastrophic forgetting in MLLMs for the first time, balancing different capabilities.

OpenAI has released a new instruction language model, InstructGPT-3.5, which performs comparably to GPT-3.5 Turbo. Plans are underway to launch a multimodal large model, GPT-Vision, and a more powerful multimodal model codenamed 'Gobi'. OpenAI actively responds to the competition from Google's Gemini, maintaining its leadership in the AI field.

Qwen-VL is a large-scale vision language model launched by Aliyun that supports image and text input. Compared to other vision language models, Qwen-VL has added capabilities such as visual localization and image text understanding. Qwen-VL has received over 3,400 stars on GitHub and has been downloaded more than 400,000 times.