AI

MokA introduces a novel, efficient multimodal fine-tuning approach that balances unimodal and cross-modal learning, boosting performance across diverse tasks and models.

A 21-year-old innovator claims his startup Cluely is transforming 9 industries with AI-powered life hacks, despite being expelled from Columbia and Harvard for controversial tools.

OpenAI begins leasing Google TPUs for ChatGPT, signaling a shift from Nvidia dominance and potential collaboration with a major rival, highlighting evolving AI infrastructure dynamics.

Anthropic's experiment with Claude managing an automated shop reveals AI's potential and risks, including hallucinations and self-awareness, highlighting future challenges in AI autonomy.

The SPACE model introduces a biologically inspired, MoE-based architecture for DNA modeling, achieving superior cross-species generalization and state-of-the-art performance in genomic tasks.

Keling AI introduces advanced multimodal video audio generation, enabling AI to produce synchronized, high-quality soundtracks and effects directly from video and text inputs, now available to users.

Mercury, a diffusion-based LLM, achieves unprecedented speed, completing tasks in seconds—over 7 times faster than Gemini 2.5 Flash, demonstrating a new era of ultra-fast AI models.

Caiyun Tech and Beijing University of Posts and Telecommunications introduce MUDDFormer, a novel architecture that enhances Transformer performance by overcoming residual connection limitations, showcased at ICML 2025.

As WAIC 2025 approaches, Shanghai announces five key advancements, highlighting global collaboration, industry integration, youth innovation, and smart city visions for the future of AI.

Researchers from Sheffield and AstraZeneca developed MapDiff, a diffusion-based model that accurately captures protein inverse folding structures and residue interactions, outperforming existing methods.