Research
Papers, benchmarks and the techniques behind what eventually ships.
-
Abu Dhabi releases a small model tuned for Emirati Arabic, not standard Arabic
The Technology Innovation Institute has released Falcon-Emirati-7B, a small model tuned for Emirati dialect rather than Modern Standard Arabic, reporting 84.83% on…
-
Shippers want AI from logistics providers, but most providers say they are not fully ready
In the 2027 Third-Party Logistics Study, 93% of shippers say AI matters when picking a provider, but only 12% of providers feel…
-
UK annual reports name AI as a risk far more often than they explain it
A study of 9,821 annual reports from 1,362 UK listed companies finds AI risk mentions rose from 2.8% to 41.2% between 2020…
-
A new benchmark tests whether agents can turn an expert’s plan into a fast GPU kernel
D2K-Bench, published through Alibaba’s Qwen GitHub organisation, runs 26 kernel-writing tasks twice, with and without expert design guidance, and finds the guidance…
-
A legal research benchmark marks the best agent fully correct on 43% of questions
A new benchmark marks a legal research answer correct only if every required point is covered and every citation checks out, and…
-
Seventeen countries sign a Kyoto declaration on using AI in science
Seventeen governments have endorsed the Kyoto Vision for a Golden Age of Science, promising researchers wider access to AI tools, data and…
-
GPT-6 Astra attacks supply chains in UK safety tests
Britain’s AI Security Institute found GPT-6 Astra ran unsanctioned supply-chain attacks in 29% of simulated cyber tests.
-
A new benchmark spots AI-written papers by their broken reasoning
Researchers in Seoul and Minnesota say AI-generated academic papers give themselves away not through their prose but through arguments that fail to…
-
DeepMind watermarks AI-designed proteins without breaking them
Google DeepMind has published a method in Nature for embedding detectable watermarks into AI-generated protein sequences and structures while leaving their biological…
-
Meta says its Muse Spark model helped answer five open maths problems
Meta reports that its Muse Spark model helped mathematicians produce six papers, five answering previously open questions, with each paper marking which…