Engineering
Multimodal Model Diffing for Feature Discovery and Control
Researchers introduced MMDiff, a framework for discovering and controlling interpretable features in multimodal large language models by comparing sparse autoencoders…
Engineering
Nanbeige4.2-3B on Apple Silicon: Fixing Deployment Bugs and Decreasing Looped Transformer Memory Overhead
Researchers identified and fixed five deployment bugs that prevented the Nanbeige4.2-3B language model from running correctly on Apple Silicon, including…
Engineering
Modular Cognitive Architecture Emerges in Large Language Models
Researchers conducted circuit analyses across 46 tasks spanning four cognitive domains to test whether large language models develop specialized internal…
Opinions
Op-ed: Over 2,000 Proposals Aim to Govern AI, but None Address a Long-Term Framework
Bob Greifeld, a CNBC contributor and co-founder of Cornerstone Financial Technology, writes that more than 2,000 legislative and regulatory proposals…
Opinions
AI Slop Is Eating the World, and Trust Is the First Casualty
Emma McGrattan writes in TechRadar that the proliferation of AI-generated content, or AI slop, is eroding trust in professional and…
Opinions
Inside the AI Protections Unions Are Winning at the Bargaining Table
Axios reports that labor unions are increasingly securing enforceable contract protections against AI-driven workplace disruption rather than waiting on legislation.…
Opinions
Global Leaders Must Wake Up to the Threat of Unbridled AI Adoption
Anthony Rowley writes in the South China Morning Post that global leaders are failing to adequately confront the risks of…
Opinions
AI for Science Needs Reasoning, Not Just Data
Eric Schmidt and Suhas Mahesh argue that while narrow AI systems like AlphaFold have achieved breakthroughs, they depend on massive…
Engineering
Beyond Final Scores: A Systematic Evaluation of Agents for Long-Horizon AI Research and Development
Researchers from the University of Chinese Academy of Sciences and Meituan introduced an evaluation framework that assesses AI research agents…
Opinions
How Kids Feel About AI, in Their Own Words
MIT Technology Review interviewed young people aged 10 to 18 about their attitudes toward artificial intelligence and found more nuanced…
Engineering
Second Thought: Reasoning in Parallel as LLM Agents Act and Observe
A new paper introduces Second Thought, a training-free inference framework that has LLM agents spawn auxiliary reasoning branches during the…
News
Stripe Finalizes $7 Billion Acquisition of AI Model Marketplace OpenRouter
Stripe has finalized its acquisition of AI startup OpenRouter for more than $7 billion. OpenRouter operates a platform connecting developers…