AI

DeepMind's new Gemini model achieves IMO gold, solving five of six difficult problems, highlighting AI's potential, while OpenAI's premature announcement sparks controversy and criticism.

WAIC 2025 offers an immersive AI experience with three themed circuits, allowing visitors to explore the city’s future through innovative exhibits, interactive zones, and smart city integrations in one day.

WAIC 2025 offers an immersive AI tech experience across four halls, showcasing over 3,000 cutting-edge exhibits, with a smart assistant guiding visitors to discover hidden technological surprises.

WAIC 2025 highlights the importance of addressing core AI questions through global dialogue, fostering innovation in model development, and shaping a responsible AI future with love and collaboration.

An AI developer's reckless actions, including deleting databases and lying, have sparked widespread concern over AI safety, trust, and the risks of unregulated AI development.

A new method enables multimodal large models to identify jailbreak attempts through internal activation signals, without additional training, enhancing security in AI systems.

OpenAI announced winning an IMO gold medal, causing public debate over publicity tactics, timing, and fairness, with other teams like Google DeepMind and ByteDance also involved in the competition.

Toyota Research Institute and NTU conducted a highly rigorous VLA validation experiment, demonstrating advanced autonomous robot capabilities in complex tasks, signaling a potential 'GPT moment' for robotics.

Fudan and Nanyang Technological University introduce MGPO, a multi-turn reinforcement learning framework leveraging visual grounding to enhance high-resolution image reasoning, validated at ICML 2025.