Bilgisayar Bilimi

Computer Science (openreview.net):
Many popular LLMs (AI models) are unable to tell the time from images of an analog clock. They are unable to answer simple calendar-based queries as well given the calendar images.
Computer Science (sciencedirect.com):
From knowledge generation to knowledge verification: examining the biomedical generative capabilities of ChatGPT
Computer Science (doi.org):
A new study finds that AI cannot predict the stock market. AI models often give misleading results. Even smarter models struggle with real-world stock chaos.
Computer Science (techxplore.com):
Groups of AI agents spontaneously form their own social norms without human help, study suggests
Computer Science (engineering.princeton.edu):
"Shallow safety alignment," a weakness in Large Language Models, allows users to bypass guardrails and elicit directions for malicious uses, like hacking government databases and stealing from charities, study finds.
Computer Science (uu.nl):
Most leading AI chatbots exaggerate science findings. Up to 73% of large language models (LLMs) produce inaccurate conclusions. Study tested 10 of the most prominent LLMs, including ChatGPT, DeepSeek, Claude, and LLaMA. Newer AI models, like ChatGPT-4o and DeepSeek, performed worse than older ones.
Computer Science (dl.acm.org):
Study investigates motivational foundations of digital self control in social media: Both high and low personal investment lead to regret and negative outcomes. Best: know why you use it but maintain personal distance.
Computer Science (doi.org):
TextRegress: A Python package for advanced regression analysis on long-form text data
Computer Science (news.uga.edu):
AI may speed up the grading process for teachers
Computer Science (pubs.acs.org):
Long-Range Resonant Charge Transport through Open-ShellDonor−Acceptor Macromolecules