Meta Launches System2Attention Mechanism to Enhance Language Model Inference Capabilities

DeepSeek and Tsinghua University Joint Research: Innovative Reward Model Inference Method Improves Scalability

Researchers from DeepSeek and Tsinghua University recently published a new paper exploring scaling methods for reward model inference, seemingly advancing DeepSeek R2. Reinforcement learning is widely used in the large-scale post-training phase of large language models, but faces the challenge of obtaining accurate reward signals for these models. The researchers found that using pointwise generative reward modeling (GRM) improves model adaptability and scalability during inference. To this end, they propose Self-Principle Calibration Tuning (SPCT) learning.

Meta's High-End Smart Glasses, Hypernova, Leaked: Features Built-in Display, Potentially $1400 Price Tag

Bloomberg reports that Meta is preparing to launch a high-end version of its Ray-Ban Meta smart glasses with a built-in display, potentially as early as the end of this year. Codenamed Hypernova, these glasses will support running apps and displaying photos, controlled via gestures and capacitive touch on the sides of the frames. The screen, according to the report, will only appear in the bottom-right quadrant of the right lens, optimized for viewing when looking downward. Upon startup, the main screen displays icons horizontally, similar to Meta's...

AliExpress March Promotion Sees AR/VR Glasses Sales Surge 600%, Brand Globalization Plan Helps Chinese Brands Break Through

Sales of AR/VR glasses on AliExpress skyrocketed by 600% during March's big promotion. A strategic brand globalization plan is credited with helping Chinese brands achieve this significant breakthrough in the international market.

GPT-4.5 Passes Turing Test with Persona: AI Conversational Abilities Reach New Heights

A recent study led by the Department of Cognitive Science at the University of California, San Diego, has achieved a landmark breakthrough in artificial intelligence: OpenAI's latest model, GPT-4.5, has demonstrated superior performance to humans in a standard Turing test using a persona, becoming the current AI system with the most human-like conversational abilities. This achievement not only reshapes our understanding of AI's language capabilities but also unlocks new potential for AI applications in social intelligence. The experiment compared four representative AI systems.

Meta's VP of AI Research, Joelle Pineau, to Depart

Joelle Pineau, Meta's Vice President of AI Research, announced on Tuesday via Facebook that she plans to leave the company in May. A highly respected leader within Meta's AI research lab (FAIR) for over two years, she oversaw the company's innovation and development in AI. Pineau's departure comes at a crucial time for Meta, as the company plans to invest up to $65 billion in AI by 2025.

Meta Unveils MoCha: AI System Transforms Text into Vivid Animated Characters with Natural Lip Sync and Movement

Meta, in collaboration with researchers from the University of Waterloo, recently launched MoCha, a novel AI system that generates full-body animated characters with synchronized speech and natural movements from simple text descriptions. This innovative technology promises to significantly enhance content creation efficiency and expressiveness, showcasing immense potential across various fields. Breaking the mold, MoCha's full-body animation with precise lip-sync differs from previous AI models that primarily focused on facial expressions. Its unique ability to render natural full-body movement makes it a groundbreaking advancement.

Meta Unveils MoCha AI System: Generating Character Animations with Synchronized Speech and Movement

Meta, in collaboration with researchers from the University of Waterloo, has developed MoCha, an AI system capable of generating full-body character animations with synchronized speech and natural movements. Unlike previous models focused solely on facial animation, MoCha renders full-body movements from multiple camera angles, encompassing lip-sync, hand gestures, and interactions between multiple characters. MoCha's demonstration highlights the accurate generation of synchronized upper body movements and hand gestures in close-up and medium shots, significantly improving lip-sync accuracy.

Meta AI Research Lead Joelle Pineau to Depart; $65B Investment Plan Continues

Joelle Pineau, VP of Meta AI Research, announced on Tuesday via a Facebook post that she will be leaving her position at Meta in May. Pineau has served as the head of Meta's Fundamental AI Research (FAIR) lab for the past two years, leading the lab's cutting-edge research in artificial intelligence. FAIR is a core internal research team at Meta, led by renowned scientist Yann LeCun. Pineau's departure comes as Meta...

Leading Authors Urge UK Government to Hold Meta Accountable for Copyright Infringement

A group of prominent authors, including Richard Osman, Kazuo Ishiguro, Kate Mosse, and Val McDermid, have signed an open letter calling on the UK government to hold Meta accountable for using copyrighted books in its AI training. The letter requests that the Secretary of State for Culture, Media and Sport, Lisa Nandy, summon Meta representatives to discuss the matter.

NVIDIA AI Researchers Introduce FFN Fusion Technology: Accelerating Large Language Model Inference