Recent advancements in Large Language Models (LLMs) have been highlighted by multiple research initiatives. Meta researchers have introduced 'System 2 distillation,' a technique aimed at improving LLM performance in complex reasoning tasks. This method fine-tunes an LLM on its own outputs, enhancing its reasoning capabilities, according to a Meta study reported by VentureBeat. Additionally, DiscoveryBench has been introduced as a comprehensive LLM benchmark to formalize the multi-step process of data-driven scientific discovery. Other notable developments include the InternLM-XComposer-2.5, a versatile vision-language model supporting long-contextual input and output, and MobileLLM, optimized for on-device use with under a billion parameters. These advancements underscore the rapid evolution and diverse applications of LLMs in AI.