Jiahui Geng

Assistant Professor

I am a tenure-track assistant professor. My research interests center on the safety, trustworthiness, and alignment of LLMs, as well as agentic AI. I aim to build AI that can reason and collaborate in software engineering and embodied tasks.

Presentation

My current research focuses on LLM (Large Language Model) safety, agentic AI for software engineering, and embodied tasks, with particular interest in robot learning. I welcome students interested in AI research to contact me, especially students from Linköping University, for discussion and potential publications.

Publications

2026

Jiaqi Li, Yun Liu, Ronghao Yu, Chuanyi Zhang, Guilin Qi, Sheng Bi, Fan Liu, Jiahui Geng, Fakhri Karray (2026)

Forgotten Horizons in Concept Erasure: Safeguarding Close-Proximity Concepts in Text-to-Image Models IEEE Transactions on Information Forensics and Security, Vol. 21, p. 7202-7214 (Article in journal)

https://dx.doi.org/10.1109/TIFS.2026.3714133

Jiaqi Li, Zihan You, Ruoyan Shen, Shenyu Zhang, Songlin Zhai, Yongrui Chen, Chuanyi Zhang, Jiahui Geng, Fakhri Karray, Sheng Bi, Guilin Qi (2026)

知识外化：多模态大型语言模型中的可逆遗忘和模块化检索第十四届国际学习表征会议 (Conference paper)

Jiahui Geng, Qing Li, Fengyu Cai, Fakhri Karray (2026)

CodeMMR: Bridging Natural Language, Code, and Image for Unified Retrieval Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (Conference paper)

Lecheng Yan, Ruizhe Li, Guanhua Chen, Qing Li, Jiahui Geng, Wenxi Li, Longyue Wang, Chenyang Lyu (2026)

虚假奖励悖论：从机制上理解 RLVR 如何激活 LLM 中的记忆捷径第43届国际机器学习大会 (Conference paper)

Fei Yang, Xuanfan Ni, Renyi Yang, Jiahui Geng, Qing Li, Chenyang Lyu, Yichao Du, Longyue Wang, Weihua Luo, Kaifu Zhang (2026)

LongSpeech：用于长语音转录、翻译和理解的可扩展基准测试 [LongSpeech: A Scalable Benchmark for Transcription, Translation and Understanding in Long Speech] ICASSP 2026 - 2026 年 IEEE 国际声学、语音和信号处理会议 (ICASSP), p. 15547-15551 (Conference paper)

https://dx.doi.org/10.1109/ICASSP55912.2026.11461821

Con Instruction: Universal Jailbreaking of Multimodal Large Language Models via Non-Textual Modalities
DOI: 10.18653/v1/2025.acl-long.146

Marco-Bench-MIF: On Multilingual Instruction-Following Capability of Large Language
DOI: 10.18653/v1/2025.acl-long.1172

VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration
DOI: 10.18653/v1/2025.findings-acl.158

Shaping the Safety Boundaries: Understanding and Defending Against Jailbreaks in Large Language Models
DOI: 10.18653/v1/2025.acl-long.1233

HD-NDEs: Neural Differential Equations for Hallucination Detection in LLMs
DOI: 10.18653/v1/2025.acl-long.309

Research

robot dog Spot, drone, small helicopter, operator

WASP at Department of Computer and Information Science

WASP, Wallenberg AI autonomous systems and software program, is the largest individual research investment in Sweden in modern times. One of the WASP research environments at LiU is located at the Department of Computer and Information Science.

TrustLLM - trustworthy, open and sustainable language model for European languages

TrustLLM is a European project coordinated by Linköping University. The project will develop a language model for European languages including Swedish and is to benefit citizens and industry alike.

About the division

The robotdog Spot hands over a first aid kit to a kneeling man

Artificial Intelligence and Integrated Computer Systems (AIICS)

The AIICS division is part of the Department of Computer and Information Science. The division's main focus is research and teaching in artificial intelligence, its theoretical foundations and its applications.

Colleagues at AIICS