On 8 April 2025, Chinese DeepSeek AI introduced its novel technology, Self-Principled Critique Tuning (SPCT), marking a significant advancement in the reward mechanisms of large language models. SPCT is designed to enhance AI models’ performance in handling open-ended, complex tasks, particularly in scenarios requiring nuanced interpretation of context and user
In a study published on 8 April 2025, researchers from Semmelweis University demonstrated that OpenAI’s GPT-4o model achieved a 93% accuracy rate in identifying acne and rosacea, while Google’s Gemini Flash 2.0 model correctly identified these skin conditions in only 21% of cases. The scientific study used
There is currently insufficient data on the impact of artificial intelligence on elections to draw well-founded conclusions, while initial threat predictions have proven exaggerated. Researchers from the NYU Center for Social Media and Politics identified only 71 instances of AI use in election-related communication in 2024. Purdue University researchers documented
Hungarian language technology research has reached a significant milestone: a comprehensive study has revealed that a larger corpus size does not necessarily lead to improved performance in morphological analysis. In their study, Andrea Dömötör, Balázs Indig, and Dávid Márk Nemeskey conducted a detailed analysis of three Hungarian-language corpora of varying
Tibor Szécsényi and Nándor Virág, researchers at the University of Szeged, have explored the context sensitivity of the huBERT language model in pragmatic annotation, focusing in particular on the automatic identification of imperative verb functions. Their study, conducted on the MedCollect corpus—a dataset of health-related misinformation—investigates how both
Researchers at Stanford University unveiled POPPER on 20th February 2025, an automated AI framework that revolutionises hypothesis validation and accelerates scientific discoveries tenfold. Following Karl Popper's principle of falsifiability, POPPER (Automated Hypothesis Validation with Agentic Sequential Falsifications) employs two specialised AI agents: the experiment design agent and the
Google DeepMind's AlphaGeometry2 artificial intelligence system has achieved a significant breakthrough in solving Mathematical Olympiad problems, performing with an 84% success rate on geometry problems from 2000 to 2024, surpassing the performance of average gold medal students. The results were published by the research team in February 2025,
Meta researchers have achieved a breakthrough in the brain-computer interface field: they have developed a new tool capable of converting brain signals into text. In the research published in February 2025, they examined the brain's language production using magnetoencephalography (MEG) and electroencephalography (EEG) with the participation of 35
The development of large language models has accelerated in recent months. The HUN-REN Hungarian Research Centre for Linguistics has now introduced the PULI LlumiX model, which is a system optimised for the Hungarian language, continuously pre-trained and fine-tuned for instruction following. A detailed presentation of the model can be found
Researchers from Stanford University, the University of Washington, and the Allen AI Institute have developed a new method to increase artificial intelligence efficiency. The s1 model, which was created with less than $50 worth of computing resources, achieves performance that was previously only possible in projects with significant budgets. The
On February 19, 2025, Google introduced a system called AI co-scientist built on Gemini 2.0, which provides virtual support for scientific researchers in developing new hypotheses and formulating research proposals. This multi-agent AI system not only creates literature reviews or in-depth analyses but is also capable of generating original
Python is one of today's most popular programming languages, yet few bug repositories contain actual, reproducible bugs. PyBugHive fills this gap as the first manually validated database containing reproducible Python bugs. The first version of the database contains 149 bugs from 11 open-source projects, providing researchers with the
