Hello, I'm Estelle Zheng

I am a first year Ph.D candidate in the SYNALP team at LORIA under the supervision of Christophe Cerisara. Prior to LORIA, I received my Master’s degree in Applied Mathematics for Data Science from Institut Polytechnique de Paris.

My research interests lies in Natural Language Processing (NLP) and more specifically in Large Language Models (LLM). My long-term research goal is to build reliable AI systems to assist and enhance decision-making across various domains.

Motivated by this goal, my current research focuses on LLM Agents and a focus on the planning stage.

Email CV Github

News

Education

CNRS - Université de Lorraine, France May 2024 - Present
Ph.D in Computer Science, advised by Christophe Cerisara. Collaborating with Alcatel-Lucent Enterprise
Institut Polytechnique de Paris, France Sept 2021 - Oct 2023
MSc in Applied Mathematics for Data Science.
Sorbonne University, France Sept 2018 - Aug 2021
BSc Double Degree in EE and Chinese Language and Culture.

Publications

Oui à l'Échelle, Non à la Mémoire : Affinage Léger des LLMs par Réseaux Latéraux

Estelle Zheng, Sébastien Warichet, Emmanuel Helbert, Christophe Cerisara
TALN, 2026 Best Paper

We revisit Ladder Side-Tuning, a parameter-efficient fine-tuning method, showing it matches QLoRA's scaling behavior while using far less peak GPU memory. We also introduce xLadder, a deeper variant with shorter reasoning chains that enables fine-tuning LLMs on consumer-grade GPUs.

Paper Preprint

Teaching

Large Language Models (Practical) 20 hours, 2025-2026
IDMC - Université de Lorraine, France Graduate Course (M2)
This course covers the fundamentals of large language models (LLMs), including their architecture (transformers), scaling laws, training methods, and applications.
Reasoning AI - Part 2 10 hours, 2026-2027
IDMC - Université de Lorraine, France Graduate Course (M2)
This course focuses on advanced reasoning techniques in AI / LLMs, including reinforcement learning, post-training methods, and agents.
Large Language Models (Practical) 20 hours, 2026-2027
IDMC - Université de Lorraine, France Graduate Course (M2)
This course covers the fundamentals of large language models (LLMs), including their architecture (transformers), scaling laws, training methods, and applications.