Renee Jia

Renee Jia

Hey there! I’m Renee. I think a lot about how intelligent systems remember, reason, and change over time — especially when the history is long and the user is a real person.

What I Do

Most of my work sits somewhere between memory, personalization, and the slightly uncomfortable question of whether a model is doing what we think it’s doing.

  • Long-horizon memory: how a system keeps anything useful from a long stretch of interaction — architectures, long-context modeling, compression and retrieval, representations that don’t evaporate
  • Personalization & recommendation: sequence learning, self-supervised user modeling, graphs, and the practical problem of doing this at scale
  • Reasoning & model behavior: chain-of-thought, faithfulness, reward hacking, shortcut learning, and trying to evaluate what models are actually up to
  • Post-training & agents: reinforcement learning, reward modeling, adaptation, and how agents behave when the task doesn’t end in one turn

Underneath all of that, I want systems that can accumulate experience, know someone a little, reason somewhat reliably, and adapt through interaction. I also care about the jump from a clean research question to something that has to live in a large production system — same ideas, just with more users and less patience.

Background

I’m currently a Research Engineer at Meta, working on long-horizon user memory, personalization, and large-scale sequence modeling.

Before that I worked at Google, on personalization and getting models into production. Even earlier, I was at Amazon Alexa AI. I’ve also spent time on research at Georgia Tech, Waterloo, and Harvard.

Publications

2026

  • From Decorative to Load-Bearing: Task Difficulty Shapes the Causal Role of Chain-of-Thought Renee Jia, Di Mu Transactions on Machine Learning Research (TMLR), 08/2026  ·  accepted
  • Correct Labels Are Not Enough: Reliable Metadata Can Silently Control Model Behavior Renee Jia, Jinzhou Wu NeurIPS 2026 Workshop  ·  under review

2023

  • Story Shaping: Teaching Agents Human-like Behavior with Stories X. Peng, C. Cui, W. Zhou, Renee Jia, M. Riedl AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE)

2022

  • Situated Dialogue Learning through Procedural Environment Generation P. Ammanabrolu, Renee Jia, M. Riedl Association for Computational Linguistics (ACL)

Reviewer: ACL, EMNLP, EACL, NAACL, CoNLL.

Let’s Connect!

Beyond the Code

When I’m not thinking about AI models, you might find me:

  • Teaching snowboarding as a PSIA-AASI Level 1 instructor
  • Playing tournament poker (check out my results at Hendon Mob)

Always excited to chat about AI research and share ideas! Feel free to reach out.