Adel Bibi, a KAUST alumnus and researcher at the University of Oxford, presented his research on AI safety, covering robustness, alignment, and fairness of LLMs. The research addresses challenges in AI systems, alignment issues, and fairness across languages in common tokenizers. Bibi's work includes instruction prefix tuning and its theoretical limitations towards alignment. Why it matters: This research from a leading researcher highlights the importance of addressing safety concerns in LLMs, particularly regarding alignment and fairness in the Arabic language.
KAUST is gradually reopening its campus after a period of lockdown, following the Saudi government's lifting of the curfew. The reopening plan incorporates best practices learned from universities worldwide and considers the evolving higher education and research landscape. KAUST has implemented comprehensive COVID-19 health and safety procedures across various aspects of life on campus. Why it matters: This measured reopening signals a return to normalcy for research and academic activities at KAUST, while prioritizing the health and safety of its community.
The Coastal and Marine Resources (CMR) Core Lab at KAUST has received two safety awards from the Royal Society for the Prevention of Accidents (RoSPA). They received a Gold Award for overall health and safety and a Bronze Award for fleet management safety. The CMR Core Lab operates a fleet of research and support vessels, including Saudi Arabia’s first fully equipped research vessel, the RV Thuwal. Why it matters: These awards highlight KAUST's commitment to safety and excellence in marine science research and operations within the region.
The paper introduces ILION, a deterministic execution gate designed to ensure the safety of autonomous AI agents by classifying proposed actions as either BLOCK or ALLOW. ILION uses a five-component cascade architecture that operates without statistical training, API dependencies, or labeled data. Evaluation against existing text-safety infrastructures demonstrates ILION's superior performance in preventing unauthorized actions, achieving an F1 score of 0.8515 with sub-millisecond latency.
Researchers introduce UnsafeChain, a new safety alignment dataset designed to improve the safety of large reasoning models (LRMs) by focusing on 'hard prompts' that elicit harmful outputs. The dataset identifies and corrects unsafe completions into safe responses, exposing models to unsafe behaviors and guiding their correction. Fine-tuning LRMs on UnsafeChain demonstrates enhanced safety and preservation of general reasoning ability compared to existing datasets like SafeChain and STAR-1.
KAUST is increasing campus population due to repatriation flights and additional students coming to campus. There has been a noticeable uptick in new cases of COVID-19, with some presenting with symptoms. KAUST emphasizes the importance of wearing face coverings, observing physical distance, washing hands, avoiding groups of more than 10 people and restricting social networks. Why it matters: This update provides insight into the university's health and safety protocols, reflecting broader trends in managing public health within research institutions in the GCC.