Yue Yu

FAIR CodeGen Team, collaborating with TBD Lab (coding squad) on end-to-end post-training (SFT/RL/eval) for Meta's next-generation LLMs, focusing on repository-level code editing, browsing-based question answering, general multi-turn usability and rubric-based RL. Core contributor of the Muse Spark model.

Meta Superintelligence Labs & GenAI Llama Research Core contributor to projects: 1) RL for Code Reasoning, 2) RL for Generalizable Coding, 3) RL for Code Testcase Generation, 4) Thinking SFT Data Generation for Helpfulness/Instruction Following on Reasoning Models, 5) RL with Rubrics-based Judge for Helpfulness/Instruction Following

• Conducted research on natural language processing and language models. • Published over 30 papers in top-tier conferences like ACL, EMNLP, and NeurIPS. • Thesis topic: Towards efficiently utilizing pre-trained language models through a data-centric lens. • Over 2000 citations as of graduation (Dec. 2024).

Generative AI Llama Post-training Team Worked on Generating Synthetic Reasoning Traces for Improving Generative Reward Models, with one paper published in NAACL 2025 Main Conference (https://arxiv.org/pdf/2411.16646).

Applied Deep Learning Research Team. Worked on Large Language Model Instruction Finetuning for Retrieval-augmented Generation with one paper published in NeurIPS 2024 (top 1% score among 15000+ submissions).

Pointer Team (now in Google Deepmind), worked on Large Language Model In-context Learning with Explanation with one paper published in ACL 2024.

Worked on Zero-shot Dense Text Retrieval with one paper published in EMNLP 2022.

Conducting Research mainly on the following topics: Spatio-Temporal Data Mining, Urban Computing, Recommender Systems (with social factors)

Outstanding Graduate Student in Tsinghua University & Beijing City (2019, top 63 of 3292 graduate students)