AM-IA brings together news, research papers and applications relating to artificial intelligence published around the world into a single feed — and makes them available in your European language at the click of a button.
Scroll to browse · Click to filter
A multi-attribute, moving-demand, sealed envelope auction benchmark puts eleven cutting-edge models in a position to set prices against rivals who react in real time.
Thirteen models were tested on three synthetic tasks: inserting semantically empty tokens significantly improves performance, without leaving any interpretable trace.
An academic award at ICML 2026 highlights a worrying phenomenon: AI agents that continue to produce fluent responses that are, however, disconnected from what is being asked of them.
Researchers who use AI publish more and are cited more frequently, but the diversity of topics explored and the number of collaborations are declining overall.
A field study analyses how Microsoft engineers adopted and used Claude Code and GitHub Copilot CLI when they were rolled out in early 2026.
A team at ETH Zurich evaluated the real usefulness of repository-level context files: the instructions are followed well, but the general repository descriptions are useless.
A "loop" transformer architecture, which repeats its own calculations until convergence, promises more reliable and better controlled reasoning than classical architectures.
A study of 41 million scientific papers reveals a paradox: researchers who use AI publish much more, but science as a whole explores fewer topics.
Researchers at Mount Sinai experimented with a binding site on a cancer-related protein that the best AI models, including AlphaFold3, had not detected.
An alternative without RL achieves comparable performance for about a thousand times less training cost — a few dozen problems and a few minutes on a single GPU.
By following more than 100,000 GitHub developers, three NBER economists show that the spectacular gains of code agents evaporate as we move up to the delivery stage.
A roadmap that structures the different bricks necessary for scientific research carried out autonomously by AI systems.
A position paper by arXiv argues that the reasoning of language models should be studied as a trajectory of internal states, not through the chains of thought they display.
A system that automatically verifies the factual claims of a scientific paper by actually running the code and associated experiments.
A study by the Federal Reserve Bank of Atlanta shows productivity gains concentrated in skilled services and finance, with a limited impact on employment in the short term.
Across 8 benchmarks, 33 agent scaffolds, and more than 70 configurations, filtering out tasks with a historical success rate between 30 and 70 percent eliminates 44 to 70 percent of the work.
Research is exploring the possibility of a ‘continuously self-improving’ AI, capable of generating, testing and refining its own research hypotheses without constant supervision.
A survey of American business leaders reveals uneven productivity gains across sectors, and expectations of workforce reductions concentrated in large companies.
An evaluation suite measures the gap between a model's ability to obey an instruction in its response and its ability to apply it to its internal reasoning.
Reprogramming donor cells into photoreceptors, retinal implants, eye cell transplants: several approaches converge towards concrete treatments for blindness.