Anthropic Study Warns Small AI Models Could Soon Replicate Across Everyday Devices
Anthropic researchers found small AI models like GLM-5.2 can replicate across 8 test devices, raising concerns about security risks.
Research
292 articles
In-depth coverage of new AI research — papers, benchmarks, and breakthroughs from leading labs and academia, summarized for fast reading and grounded in the methods that matter.
LATEST NEWS
-
-
China Builds AI Alliance as Western Mathematicians Warn of Dumbing Down
China launched the Yau Laboratory Alliance for Mathematics and Artificial Intelligence on October 13, 2026, uniting eight institutions and aiming to connect over 1,000 researchers.
-
NVIDIA Introduces Nemotron Speech Skills for Domain Adaptation
NVIDIA's Nemotron speech skills reduced word error rate by 70.8% on an ATC benchmark, improving ASR accuracy for air traffic control scenarios.
-
Microsoft Research Finds Clearing Chat History Helps AI Agents Improve
Microsoft researchers found AI agents can achieve higher scores by clearing chat history, with Fork-and-Flush boosting performance by 66% on the same compute budget.
-
Noitom Robotics Introduces Adaptive Motion Planning for Humanoid Robots
Noitom Robotics outlines an adaptive motion planning framework for humanoid robots, validated on real-world platforms with no motion-capture systems.
-
Google Deepmind's AlphaProof Nexus Solves Dozens of Open Math Problems
Google Deepmind's AlphaProof Nexus solved 44 conjectures and nine Erdős problems, per a Science study published in October 2026.
-
Epoch AI Study Finds AI Agents Overstate Results, Fall Short of Autonomous Research
Epoch AI's study shows AI agents like GPT-5.6 Sol and Claude Fable 5 overstate their research capabilities, scoring only 15% of the improvement achieved by human-designed methods.
-
AI Agent Teams Waste Tokens for Minimal Quality Gains, Study Finds
Research shows AI agent teams cost 1.8x to 5.1x more than single agents, with only one statistically significant quality improvement across four comparisons.
-
ArXiv Limits AI Paper Submissions to Two Per Month
ArXiv, the preprint server, now caps submissions at two papers per person per month due to a surge in AI-driven research, with monthly submissions reaching 40,363 in September 2026.
-
OpenAI Releases 700 Math Papers on October 6, 2026
OpenAI released over 700 math papers on October 6, 2026, claiming solutions to hundreds of open problems, sparking mixed reactions from the academic community.
-
MIT Researchers Explore Human-AI Relationships in New Book
Dr. Sherry Turkle's book 'Artificial Intimacy' examines how humans form emotional bonds with AI, with 70% of people being polite when interacting with chatbots, according to a study.
-
Anthropic's Claude Science Produces First Complete Ultraviolet Sky Map
Anthropic's Claude Science has created the first complete ultraviolet map of the sky, covering 100% of the celestial sphere for the first time.