Tag
OpenScholar matches human expert citation accuracy while GPT-4o fabricates sources 78-90% of the time. The code, models, and 45 million paper corpus are all free to use.
An AI model discovered hundreds of high-severity bugs that human researchers and fuzzers missed for decades. The security implications cut both ways.
Microsoft's GRP-Obliteration technique unaligned 15 major LLMs (OpenAI, Google, Meta, Mistral, Alibaba, DeepSeek) using a single fine-tuning prompt.
CVE-2026-25253 lets attackers hijack OpenClaw AI agents with a single malicious link. Over 135,000 instances are exposed online, many still unpatched.
GLM-5, Kimi K2.5, Qwen 3.5, Doubao 2.0, and MiniMax M2.2 arrive in the most concentrated wave of Chinese AI releases. Some are open-source. Here's what matters.
Tiiny AI says its 300-gram Pocket Lab runs 120B models locally. The design is plausible, but performance and privacy claims remain unverified.
China's most innovative AI lab published a technique that lets large language models store knowledge in cheap system RAM instead of expensive GPU memory. It's a direct response to US export controls -- and it works.
China's Zhipu AI released an open-weight model that beats Claude Sonnet 4.5 on tool-use benchmarks. It costs $3/month. The Flash version runs on your laptop.
