Meta has established a secret AI research team named MRS Research to enhance its advertising algorithms for Facebook, Instagram, and Threads. The unit has operated since October 2025 and is part of Meta’s Recommendation Systems division, collaborating closely with the advertising division that accounts for nearly all of the company’s $201 billion revenue in 2025.
Yang Song leads MRS Research. He joined Meta in November 2025, previously managing recommendation algorithms at TikTok. Song has stated that the team’s mission is to transform recommendation systems through large language models, aiming to improve understanding and reasoning capabilities in ad placement and organic content.
The team has attracted talent from rival companies, including Lihong Li from Amazon, Xiaolong Wang from OpenAI, and Fei Sha from Google. This suggests a strategic move by Meta to bolster its expertise in AI-driven recommendations.
MRS Research operates independently from Meta Superintelligence Labs, which focuses on general-purpose AI. A Meta spokesperson clarified that while Superintelligence Labs pursues advanced AI models, MRS Research specifically targets the optimization of content recommendations.
The lab’s formation came into public awareness through job postings on LinkedIn, highlighting its commitment to improving content and user understanding across Meta’s platforms. The increased investment in this area accompanies Meta’s broader shift towards AI-driven ad optimization.
In late 2025, Meta introduced a new model purported to improve advertising effectiveness, complemented by a rollout of an Adaptive Ranking Model on Instagram to enhance ad-serving processes. Morgan Stanley has projected a 28% growth in Meta’s advertising revenue for 2026, indicating strong market optimism regarding these developments.
