Publications

A collection of my research work.

Algorithmic Recourse of In-Context Learning for Tabular Data

Wenshuo Dong^†, Jiaming Zhang^†, Shaopeng Fu, Hongbin Lin, Di Wang, Lijie Hu

ICML 2026

This paper presents the first study of algorithmic recourse for tabular prediction under in-context learning (ICL). It proposes ASR-ICL, a query-efficient black-box framework that generates sparse and actionable recourse for ICL-based models.

Understanding and mitigating cross-lingual privacy leakage via language-specific and universal privacy neurons

Wenshuo Dong, Qingsong Yang, Shu Yang, Lijie Hu, Meng Ding, Wanyu Lin, Tianhang Zheng, Di Wang

EMNLP 2026

This paper investigates cross-lingual privacy leakage in LLMs, showing that models trained on sensitive data in one language can still leak private information when queried in other languages.

Eap-gp: Mitigating saturation effect in gradient-based automated circuit identification

Lin Zhang, Wenshuo Dong, Zhuoran Zhang, Shu Yang, Lijie Hu, Ninghao Liu, Pan Zhou, Di Wang

NeurIPS 2025

This paper proposes EAP-GP, a gradient-based method for automated circuit identification in language models that mitigates the saturation effect in existing attribution methods.

Trend forecasting of public concern about low carbon based on comprehensive Baidu index and its relationship with CO 2 emissions: The case of China

Wenshuo Dong^†, Renhua Chen^†, Xuelin Ba, Suling Zhu

Sustainability 2023

This paper proposes a hybrid CEEMD-SGIA-LSTM forecasting framework to predict public concern about low-carbon issues in China using the Baidu Index. It further analyzes the relationship between public low-carbon awareness and CO₂ emissions, showing an overall increasing trend in public concern.