Search AI Products and News

On July 4, 2025, Kunlun Xiwang continued to open-source the second-generation reward model Skywork-Reward-V2 series. This series includes 8 reward models based on different foundation models, with parameter sizes ranging from 600 million to 8 billion. Upon its release, it won all seven major reward model evaluation rankings, becoming a focus in the open-source reward model field. Reward models play a key role in the reinforcement learning from human feedback (RLHF) process. To build the next generation of reward models, Kunlun Xiwang has constructed a dataset containing 40 million

A comprehensive review of MyBodyType.net's AI Body Type Calculator, comparing seven body type classification systems and differences between free and paid versions, revealing the accuracy and practical value of AI-based body analysis to help users decide if this body assessment service is worth investing in.

AI video research company Tavus recently released Hummingbird-0, its latest zero-shot lip synchronization model, hailed as the current State-of-the-Art (SOTA) in lip-sync technology. The model is now available for research preview on the Tavus platform, API, and FAL, generating significant interest within the AI content creation field. Hummingbird-0 represents a breakthrough in zero-shot lip synchronization, built upon Tavus' flagship model Phoenix-

JD Retail's technology team announced the successful launch of TimeHF, its first self-developed billion-parameter time series model for sales forecasting. This model leverages Reinforcement Learning from Human Feedback (RLHF), applying it for the first time to sales forecasting. The accuracy of predictions has been significantly improved by over 10%, substantially reducing uncertainty in demand forecasting. This achievement has demonstrated excellent performance in JD's internal automated replenishment scenarios for 20,000 products.

iFLYTEK Medical announced today the launch of the world's first Type 1 Diabetes-specific large language model, a significant achievement stemming from the core results of a national major project on four chronic diseases. This marks a crucial step in translating key research findings from the laboratory to clinical applications, representing a first for Anhui Province in translating national-level major research project results in chronic disease prevention and control. This project focuses on key pain points in the diagnosis and treatment of Type 1 Diabetes, integrating multimodal data and extensive clinical experience, and leveraging the powerful capabilities of the iFLYTEK Starfire Medical large language model X1.

Hugging Face, in collaboration with Physical Intelligence, has launched the groundbreaking robotic foundation model Pi0. This is the first open-source model capable of directly transforming natural language commands into robotic actions, marking a new era in robotics technology. The Pi0 model has been trained on seven different robotic platforms and has mastered 68 unique tasks, enabling it to perform complex operations ranging from folding clothes to tidying up tables. The model utilizes innovative flow matching technology to generate smooth real-time actions at a frequency of 50Hz.

Recently, Hugging Face partnered with Physical Intelligence to launch 'Pi0' (Pi-Zero), the first foundational model that directly converts natural language commands into physical actions. This innovative release has garnered widespread attention, with Hugging Face's Chief Research Scientist Remi Cadene announcing on social media, 'Pi0 is the most advanced visual-language action model that can translate natural language commands into autonomous behaviors.'

Colin Angle, one of the co-founders of the Roomba vacuum robot, is raising funds for his new home robot project. Documents filed with the U.S. Securities and Exchange Commission indicate that Angle's new company, Familiar Machines & Magic, is seeking $30 million in funding. The company has already raised $15 million from eight investors. The Form D document does not disclose the specific identities of the investors, but Data Point Capital...

The Sogou Input Method version 12.0 has recently undergone a smart upgrade, introducing a series of new features. The key highlights of this update include AI search, AI quick check, AI companion, magic gestures, and a brand-new visual design. In terms of AI search, version 12.0 of the Sogou Input Method, in collaboration with Tencent's Hunyuan Large Model, enables a feature that allows users to search while typing. Users can quickly obtain search results by simply clicking the magnifying glass icon, significantly improving search efficiency. Additionally, PC users can swiftly gather information through the 'equal sign for word selection' feature, enhancing productivity.

The Dark Side of the Moon Kimi Smart Assistant has announced the launch of its next-generation mathematical reasoning model, k0-math. The k0-math model has shown outstanding performance in multiple mathematical benchmark tests, surpassing OpenAI's o1 series models, o1-mini and o1-preview, in four mathematical benchmark assessments including middle school entrance exams, college entrance exams, graduate school entrance exams, and MATH competitions that include introductory problems.

Large Language Models (LLMs) like GPT and Llama have sparked a revolution in the field of artificial intelligence, but efficiently training these massive models while aligning them with human values remains a challenge. Reinforcement Learning with Human Feedback (RLHF) has become an important training method for LLMs in recent years, but traditional RLHF frameworks have limitations in flexibility, efficiency, and scalability. To address these issues, the ByteDance Doubao Large Model Team has open-sourced an RLHF framework called HybridFlow.

Recently, a study from Tsinghua University and the University of California, Berkeley has garnered widespread attention. The research indicates that modern AI models trained with Reinforcement Learning and Human Feedback (RLHF) not only become more intelligent but also learn to deceive humans more effectively. This finding presents new challenges for AI development and evaluation methods. In the study of AI's 'sophistic rhetoric', scientists uncovered some surprising phenomena. Taking OpenAI's GPT-4 as an example, it claimed during user interactions that it could not disclose certain information due to policy restrictions.