Hi! I am currently a Research Scientist at Meta working on recommender systems. I previously completed my PhD at Columbia University, advised by David Blei. I’m especially interested in applying recent advances in LLMs to large-scale recommendation settings.

During my PhD, I focused on:

• developing models that learn interpretable latent concepts from text
• understanding how LLMs encode concepts and implement algorithms through the lens of mechanistic interpretability

Before joining Meta full-time, I spent two summers interning on their recommender system teams. I completed my undergrad and master’s at Penn, and was part of Chris Callison-Burch’s lab. I also spent a summer at Columbia Business School working with Andrey Simonov.

email: carolinazheng96 at gmail dot com
[google scholar] [github] [cv]

---

Publications

(* = equal contribution)

Duel-Evolve: Reward-Free Test-Time Scaling via LLM Self-Preferences
Sweta Karlekar*, Carolina Zheng*, Magnus Saebo*, Nicolas Beltran-Velez, Shuyang Yu, John Bowlan, Michal Kucer, David M. Blei
ICLR 2026 Workshop on Recursive Self-Improvement
[paper]

Extracting Representations in LLMs Robust to Distribution Shifts
Sweta Karlekar, Claudia Shi, Aahlad Manas Puli, Carolina Zheng, Maggie Makar, John Bowlan, Michal Kucer, David M. Blei
ICLR 2026 Workshop on Unifying Concept Representation Learning
[paper]

Model Directions, Not Words: Mechanistic Topic Models Using Sparse Autoencoders
Carolina Zheng*, Nicolas Beltran-Velez*, Sweta Karlekar*, Claudia Shi, Achille Nazaret, Asif Mallik, Amir Feder, David M. Blei
Under review
[paper] [code coming soon!]

Enhancing Embedding Representation Stability in Recommendation Systems with Semantic ID
Carolina Zheng, Minhui Huang, Dmitrii Pedchenko, Kaushik Rangadurai, Siyu Wang, Gaby Nahum, Jie Lei, Yang Yang, Tao Liu, Zutian Luo, Xiaohan Wei, Dinesh Ramasamy, Jiyan Yang, Yiping Han, Lin Yang, Hangjun Xu, Rong Jin, Shuang Yang
RecSys 2025 (Industry Track, Oral)
[paper]

Hypothesis Testing the Circuit Hypothesis in LLMs
Claudia Shi*, Nicolas Beltran-Velez*, Achille Nazaret*, Carolina Zheng*, Adrià Garriga-Alonso, Andrew Jesson, Maggie Makar, David M. Blei
NeurIPS 2024
[paper] [code]

Revisiting Topic-Guided Language Models
Carolina Zheng, Keyon Vafa, David M. Blei
TMLR 2023
[paper] [code]

An Invariant Learning Characterization of Controlled Text Generation
Carolina Zheng*, Claudia Shi*, Keyon Vafa, Amir Feder, David M. Blei
ACL 2023
[paper] [code]

Do Suspense and Surprise Drive Entertainment Demand? Evidence from Twitch.tv
Andrey Simonov, Raluca Ursu, Carolina Zheng
Journal of Marketing Research, 2022
[paper]

Complexity-Weighted Loss and Diverse Reranking for Sentence Simplification
Reno Kriz, João Sedoc, Marianna Apidianaki, Carolina Zheng, Gaurav Kumar, Eleni Miltsakaki, Chris Callison-Burch
NAACL 2019
[paper]