IBM releases Granite Time Series PatchTST-FM-r2 for forecasting
The new time-series model is offered under a license compatible with commercial uses.
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
The new time-series model is offered under a license compatible with commercial uses.
OpenAI announces that it has demonstrated a variant of the Navier-Stokes equations; mathematician Tristan Buckmaster accuses the company of having wanted to dismiss his co-author because he works at Anthropic.
Ten thousand agents mobilized for 88 hours produced proof showing that the Navier-Stokes equations can, under certain conditions, cease to have a regular solution.
Thermodynamic constraints in models, reproducible data, inverse design and automated feedback: Tohoku University is designing an ecosystem where physics frames prediction.
Developed at NIST with the University of Maryland, the algorithm fires X-ray beams at a tin-bismuth wafer and extrapolates beyond the already measured points to decide where to look next.
A decade of work has resulted in the mapping of 166,000 neurons and 125 million synaptic connections, the largest brain map ever made.
Tested five times on the DIII-D tokamak in San Diego, the software framework at the Princeton Plasma Physics Laboratory reads and commands in about twenty milliseconds, whereas a human operator reacts in seconds.
The European Research Council is funding a facility in Durham that combines a robotic laboratory and explainable AI, with the aim of reducing the discovery of functional molecules from several decades to a few weeks.
Anima Anandkumar and Benedikt Jenik launched Accelerated Understanding, whose model abandons the Transformer in favor of neural operators and ingested 5 trillion data points in a single query.
The joint lab wants to run generative and embodied AI on energy-constrained components, relying on 18 nm FD-SOI and phase-change memory.
By exploiting all multispectral reflectance rather than chlorophyll-a alone, models estimate the pigments that make it possible to identify phytoplankton groups.
Sakana AI, UBC, the Vector Institute, and Oxford detail the system that produced the first fully AI-generated paper to pass peer review.
The analysis identifies the uses that have become common—exploration of conjectures, formal verification, generation of counterexamples—and emphasizes what automation does not replace.
A ridge regression per attention head, calibrated over 500 sequences, allows the KV cache of one model to be reused in another: from 2.7 to 25 times faster than a full recalculation.
The Translation Readiness Index scores the lexical proximity between an article abstract and publications already associated with patents. The model achieves 78% correct rankings.
By digging through the texts, tables and graphs of 448 articles, a team put together a dataset and then designed temperature-stable lead-free dielectrics by inversion.
Data scientists at the Jefferson Lab have designed a digital twin that anticipates the micro-displacements of the DIII-D tokamak's coils between two shots, with 80% less error compared to static models.
Open since 2004, this numerical linear algebra conjecture gave way to a neurosurgery resident with no training in advanced mathematics, guiding a model in autonomous operation.
A team from Seoul National University combines automatic literature extraction and physics-guided learning to design lead-free dielectrics that are stable at high temperatures.
Jason Zhang, trained in the laboratory of Nobel Prize winner David Baker, is designing new proteins by computer and testing 20,000 of them in a single experiment.