Yi Jing

Current internships

I am currently interning at THUKEG and PKUICL under the joint supervision of Prof. Juanzi Li and Prof. Zhifang Sui.

Previous research

Previously, I conducted summer research at the University of Maryland with Prof. Jordan Boyd-Graber, interned at THBI, and participated in summer research at Oxford University's FMRIB, supervised by Prof. Qiyuan Tian and Prof. Wenchuan Wu.

Opportunities

I look forward to internship and summer research opportunities in interpretability, post-training, and AI for research.

Interpretability of large language models

Investigating the principles and mechanisms underlying advanced cognitive capabilities in large language models, then using those insights to optimize model training and downstream applications.

Post-training

Studying how to enhance model performance on downstream tasks through methods such as reinforcement learning, with a particular focus on creative tasks.

AI for Research

Building AI systems for research with the long-term goal of enabling AI self-evolution.

Open source interpretability toolkit

OpenSAE: Open Source Sparse Auto-Encoders and Toolkits

Open-source sparse autoencoders and toolkits for language model interpretability.

Preprint

Guiding LLM Post-training Data Engineering with Model Internals from Sparse Autoencoders

A sparse autoencoder-guided framework for using model internals to engineer post-training data for LLM reinforcement learning.

ACL 2026 Main

HistLens: Mapping Idea Change across Concepts and Corpora

A method for mapping how ideas change across concepts and corpora.

EMNLP 2025 Main

LinguaLens: Linguistic Mechanism Analysis and Control Framework of LLMs

A framework for analyzing and controlling linguistic mechanisms in large language models.