Brief

Researchers have identified a significant vulnerability in classifier-based quality filtering, a common technique for curating pre-training data for large language models. Their study demonstrates that simple reformatting of content, mimicking Wikipedia's style, can trick these classifiers into misjudging document quality. This could lead to the inclusion of lower-quality data in training corpora, potentially impacting model performance. AI

Project N.O.M.A.D. aims to create a personal server that can run essential services like Wikipedia, offline maps, courses, and local AI applications on any computer. This initiative focuses on providing access to information and tools even when internet connectivity is lost. The project seeks to empower users with self-hosted digital resources. AI

Researchers have introduced TailedTS, a new benchmark dataset designed to evaluate time series forecasting models on data exhibiting heavy-tailed, zero-inflated, and non-Gaussian distributions. Derived from Wikipedia page view data, TailedTS contains approximately 24.69 billion data points, highlighting that a small percentage of pages receive a majority of views, thus creating a challenging testbed for model robustness. The dataset also facilitates research into periodicity quantification and standardized prediction benchmarks using non-Gaussian loss functions, revealing that standard estimators perform poorly on high-volume data. AI

A recent study indicates that AI-generated articles on Grokipedia are less readable and cite fewer sources than human-edited Wikipedia articles. The research, published in the Proceedings of the National Academy of Sciences, found that automated encyclopedias tend to produce longer, more complex content. Furthermore, these AI-generated articles may exhibit a rightward political bias in certain subject areas. AI

Researchers have developed Stage-Audit, a system designed to improve the accuracy and source-grounding of tables generated by large language models. The system addresses the issue of LLMs fabricating or misattributing sources for table entries by implementing distinct curator and auditor roles with write permissions. Stage-Audit also incorporates a row-level source-citation gate and a comprehensive audit taxonomy to ensure explicit traceability of information. AI