introduction to the internal family systems model

Tiangong was a domestic language model that was developed by Kunlun Wanwei and Qidian Zhiyuan. It was the first domestic language model in China that truly realized the emergence of intelligence. The invitation test was launched on April 17, 2023. It was a generation-based AI product after Kunlun Wanwei's AI painting product "Heavenly Craftsmanship." It could meet the needs of copywriting, knowledge question and answer, code programming, logic deduction, mathematical calculation, and so on. It had a unique advantage. It could support text conversations of up to 10000 words and could interact with users for more than 20 rounds. On July 27,2023, the model was successfully selected into the "Foundation-Building Plan" of the China Academy of Information and Communication Technology, a panoramic view of high-quality digital transformation products and services (2023). On November 3, the model was filed through the "Temporary Methods for the Management of Generative Artificial Intelligence Services" and opened to the whole society. Kunlun Wanwei began to lay out the AIGC field in 2020 and invested tens of millions of yuan to set up a research and development team to develop the Tiangong model. It went through multiple research and development stages, such as the development of the Chinese GMT- 3 model with 10 billion parameters from the end of 2020 to April 2021. On April 10th, 2023, it announced the release of version 3.5 of Tiangong. Its intelligence level was very close to that of the Open AI ChatGMT. Skywork O1 was launched on November 27, 2024. This was a series of models with complex thinking and reasoning abilities. It was the first O1 model in China with Chinese logical reasoning abilities. It supported thinking, planning, and reflection on the model output. This version includes Skywork O1 Open (An open source model based on Llama 3.18B. The evaluation index has been greatly improved to the level of SOTA in the same ecological niche open source model. It has unlocked many complex mathematical tasks that cannot be solved by lightweight models and can help accelerate the process of replicating o1 in the domestic open source community.), Skywork O1 (Complete thinking ability, better Chinese support, faster reasoning and thinking speed, outstanding performance in mathematics, Chinese logic, and reasoning problems), Skywork O1 Previews (Complete version of reasoning model, matched with self-developed online reasoning algorithm, compared to Skywork O1 Litehas a more diverse and "deep" thinking process, more complete and higher quality reasoning). It would also open source the two reasoning task's Process Reward Model (Prm): Skywork O1Open- Prm - 1.5B and Skywork O1Open- Prm -7B. The huge improvement in the performance of Tiangong Model 4.001 was due to Tiangong's three-stage self-developed training program, including reflective reasoning ability training, reasoning ability reinforcement learning, reasoning planning, etc. If you want to know more about the follow-up, click on the link and read it!

Shili was one of the more famous models in the Han clothing industry. She was a regular model for the Han Shang Hualianjia. Her style was ever-changing, and she could be called the " treasure of the shop " of Han Shang Hua Lian. There was a saying that " a smile from ten miles away can cause your wallet to collapse ", which showed how charming she was. Judging from her image and style, she could control many styles. For example, when she wore a red and white Han Dynasty dress with a paper kite in her hand, she looked cute and cute, like a young lady who sneaked out to play. The braids on her head and sandals on the soles of her feet added points to her overall appearance. When she wore a red, black and white round neck robe, she looked valiant. A high ponytail matched with a red belt on her forehead, and she looked even more handsome with a long spear in her hand. When she wore red clothes, she was shy and shy, like a young lady waiting to be married. The light blue Han Dynasty dress looked fresh and gentle, and her long hair made her look like a girl next door. The light pink and light blue Ming-made Han clothes were dignified and noble, and the exquisite hair crown made them even more elegant. The white Ming-made Han clothes were gentle and moving, and the exquisite hairstyle and hair accessories added a lot of color to the overall style. She was wearing a light green gauze dress and a dark blue chest-length skirt. When she matched it with the "Nezha head", the red and blue ribbons on her chest fluttered in the wind. She was lively and cute when playing rattan ball, smart and playful when walking, and heroic when she bent her bow and shot arrows. It could be sweet or salty, and it was suitable for both movement and stillness. "40 Sisters" is equally exciting. Everyone is welcome to click and read it!

There was a model of the Long March 5 launch vehicle launched by the Keeppley brand. This model was a genuine authorized product of China Aerospace, and it performed well in terms of proportion and details. The alloy model had a ratio of 1:150. It was a two-and-a-half-stage rocket with four boosters. It could reflect some of the characteristics of the Long March 5 rocket and could be used as an item for aerospace science or collection. It was currently sold in the 3600-yuan store. The original price was 199 yuan, and you could enjoy a 15% discount if you placed an order. The actual starting price was 169.2 yuan. The novel " Hundred Years of Spaceship " is equally exciting. Everyone is welcome to click and read it!

The hunyuan AI model was a big language model developed by tencent. It had strong Chinese writing ability, logical reasoning ability in complex context, and reliable task execution ability. Its main products were the Mixed Yuan Life Text and the Mixed Yuan Life Map, which could complete multiple rounds of questions and answers in professional fields. It could support literary creation, text abstractions, and role-playing. It could infer and analyze based on existing data or information, and at the same time, effectively solve facts and time-sensitive problems. The Grand Mixed Model had a scale of over 100 billion parameters, and the pre-training data exceeded 2 trillion tokens. On August 3,2023, it entered the internal testing stage of the application. In 2023, it was successively filed by the National Network Information Office and the "Temporary Methods for the Management of Generative Artificial Intelligence Services". On September 6th, the Mini programs "Tengyuan Assistant" was launched on WeChat. The next day, it officially released the big model. In December, it passed the domestic "big model standard compliance evaluation" test. On May 14th, 2024, it announced that it would be open source. On May 30th, the application based on the Hunyuan model,"Tengyuanbao," was launched, connecting to more than 600 Tengyuanbao business scenarios. In April 2024, the Mixed Element Grand Model technical architecture was upgraded to the Mixed Expert Model (MoE) architecture. The scale of parameters reached trillions, and it could handle complex scenes and multi-task scenes. In addition, the Mixed Yuan Life Map had a strong Chinese language comprehension ability. It could generate matching pictures according to the keyword information in the description and support functions such as word expansion.