Search AI Products and News

In recent years, the field of artificial intelligence has undergone tremendous changes, especially with large language models (LLMs) making remarkable progress in multi-modal tasks. These models have demonstrated strong potential in their ability to understand and generate language, but most current multi-modal models still adopt auto-regressive (AR) architectures, which limit inference processes to be rather monotonous and lacking flexibility. To address this limitation, a research team from The University of Hong Kong and Huawei Noah's Ark Lab has proposed a brand new model – FUDOKI, aiming to break these constraints. The core innovation of FUDOKI is

Recently, Apple Inc. released a controversial paper pointing out significant defects in the reasoning abilities of current large language models (LLMs). This view quickly sparked heated discussions on social media, especially among senior software engineers on GitHub, such as Sean Goedecke, who strongly opposed it. He believed that Apple's conclusions were too one-sided and could not fully reflect the capabilities of reasoning models. Apple's paper pointed out that the performance of LLMs was unreliable when solving benchmark tests such as mathematics and programming. The research team from Apple adopted

At the recent annual general meeting held at Meituan headquarters in Hengdian Building in Beijing, founder Wang Xing made a systematic exposition of the company's strategic layout and development plan in the field of artificial intelligence for the first time, revealing the deep thinking and bold practice of Meituan in the AI wave. Wang Xing divided Meituan's AI development into different stages. In the early stage, the company had already applied deep neural network algorithms in its food delivery route and dispatch systems. The current phase focuses on the development and deployment of large language models and their derivative applications. Facing fierce competition in the AI field, Meituan has carried out...

Recently, Meta's FAIR team and researchers from the Hebrew University of Jerusalem jointly released a new study indicating that reducing the reasoning time of large language models can significantly improve their performance in complex reasoning tasks. The research findings show that using shorter reasoning chains, the accuracy of AI models has increased by 34.5%, which challenges some assumptions in the current AI industry. Image source notes: The image was generated by AI and is licensed by Midjourney in this study, the authors point out that prolonged deliberation

Recently, Google officially released the open source framework LMEval, aimed at providing standardized evaluation tools for large language models (LLMs) and multimodal models. The launch of this framework not only simplifies cross-platform model performance comparisons, but also supports assessments in areas such as text, images, and code, showcasing Google's latest breakthroughs in the field of AI evaluations. AIbase has compiled the latest developments of LMEval and its impact on the AI industry. Standardized Evaluations: Simplified Cross-Platform Model Comparisons

Programmers have a new tool to cheer about! LocAgent, a graph-indexed large language model (LLM) agent framework designed specifically for code localization, has been jointly developed by the research teams from OpenHands, Yale University, University of Southern California, and Stanford University. Excitingly, LocAgent's code localization accuracy reached 92.7%, and this new achievement will be officially presented at the 2025 ACL conference. Programmers often struggle with identifying the root cause of code problems in their work. For example

Mistral recently launched its new Agents API, a framework designed for developers to simplify the creation of AI agents that can execute various tasks such as running Python code, generating images, and performing Retrieval-Augmented Generation (RAG). The launch of this API aims to provide a unified environment for large language models (LLMs) to interact with multiple tools and data sources in a structured and persistent manner.

At a critical moment when artificial intelligence and blockchain technology are rapidly converging, the open-source AGI research institution DMind is leading industry change with its innovative Web3-specific language models. Recently, the institution released two domain-specific large language models, DMind-1 and DMind-1-mini, which have been deeply optimized for Web3 application scenarios such as blockchain, decentralized finance, and smart contracts. The technical breakthrough in specialized AI models: the DMind-1 series of models represent an important advancement in the application of AI technology in vertical fields. Compared to traditional models...

When interacting with large language models like ChatGPT or Claude for a long time, have you ever encountered embarrassing situations where the conversation content suddenly forgets? This is not intentional on the part of AI, but rather limited by the inherent context window restrictions of large language models. Regardless of whether it's an 8k, 32k, or 128k token capacity, once this threshold is exceeded, the previous conversation content will be truncated and lost, seriously damaging the interactive experience. Recently, a company called Supermemory has launched a disruptive technology—Infin

Recently, the ModelScope community announced the release of a new dynamic benchmark dataset named UGMathBench. Its goal is to comprehensively assess the mathematical reasoning abilities of language models across a broad range of undergraduate mathematics subjects. The advent of this dataset fills the current gap in evaluating the reasoning capabilities of language models in the field of undergraduate mathematics and provides researchers with a richer and more challenging testing platform.