Qi Qi

About

I am a Senior Research Scientist at Meta Superintelligence Labs, where I work on long-context and long-horizon post-training for frontier language models. I work closely with Xinyun Chen and Denny Zhou.

I led long-context post-training for Llama 4 Maverick and have been a core contributor to the long-context capabilities of Muse Spark 1.x. Most recently, Muse Spark 1.2 achieved 83.3% SOTA performance on AA-LCR, ranking #1 across all open and closed models.

I received my Ph.D. in Computer Science from the University of Iowa, advised by Professor Tianbao Yang. My earlier research spans machine learning and optimization, including distributionally robust optimization, learning with imbalanced data, and multimodal learning.

News

• [Apr-2026] Try Meta AI with our new Muse Spark model!
• [Apr-2025] Llama4 Released, try our new model with context length up to 10M! Link Try our product! Meta AI
• [Nov-2022] Try our new package ABSGD for handling imbalanced classification! Tutorial
• [Oct-2022] Our ABSGD achieved 1st in ResNet50 (rank 4th of 16 overall) on Stanford WILDS iWildCam! Code
• [Aug-2022] Research intern at Netflix with Shervin Ardeshir and M. Hossein Taghavi
• [Jan-2022] Invited talk at UberAI about AUPRC maximization

Selected Publications

Muse Spark 1.2: State-of-the-Art Long-Context Reasoning

Core Contributor: Qi Qi

83.3% SOTA · #1 across open and closed models on AA-LCR / leaderboard / blog / demo

The Llama 4 Herd: The Beginning of a New Era of Natively Multimodal AI Innovation

AI Meta (incl. Qi Qi)

Technical Report 2025 (🔥 378+ citations) / blog / demo

Professional Service

Conference Reviewing: ICML / NeurIPS / ICLR / CVPR / IJCAI