Biography
Here is Guanhua Chen (陈冠华). I am a tenure-track assistant professor in the Department of Statistics and Data Science in Southern University of Science and Technology (SUSTech). I received Ph.D. from the Department of Computer Science, The University of Hong Kong in 2022, under the supervision of Prof. Jia Pan and Prof. Wenping Wang. I obtained the Bachelor and Master degree from Tsinghua University in 2012 and 2014, respectively. I was a research intern in Microsoft Research Asia and Huawei Noah’s Ark Lab. Currently, I am an area chair of ACL, EMNLP and CCL, also the reviewer for top AI conferences and journal like ICML, NeurIPS, ICLR, ACL, EMNLP, NAACL, and TASLP. I was awarded as Microsoft Research Asia StarTrack Scholar in 2025. My research interests are natural language processing (NLP) and large language models (LLMs), especially reasoning LLMs, multimodal LLMs, agentic RL and LLM for low-resource applications like health and engineering, etc.
Positions available: I am looking for self-motivated PostDoc/PhD/Master/visiting students to join our lab. If you are interested, please send me an email with your CV. For your reference, here are the latest information of the PhD and Master application process of this year. Currently, we have 16 RTX 4090 GPUs (24GB), 16 NVIDIA L40 GPUs (48GB), and 8 RTX Pro 5000 GPUs (72GB) available for students. Enough APIs of open-sourced/proprietary LLMs are also provided for students.
Selected Publications
（* denotes corresponding author）
SPPO: Sequence-Level PPO for Long-Horizon Reasoning Tasks
- Tianyi Wang, Yixia Li, Long Li, Yibiao Chen, Shaohan Huang, Yun Chen, Peng Li, Yang Liu, Guanhua Chen*
- In Proceedings of ACL 2026 (CCF-A, main conference)
GIFT: Guided Fine-Tuning and Transfer for Enhancing Instruction-Tuned Language Models
- Zhiwen Ruan, Yichao Du, J. Zheng, Longyue Wang, Yun Chen, Peng Li, Jinsong Su, Yang Liu, Guanhua Chen*
- In Proceedings of ACL 2026 (CCF-A, main conference)
From Word to World: Can Large Language Models be Implicit Text-based World Models?
- Yixia Li, Hongru Wang*, Jiahao Qiu, Zhenfei Yin, Dongdong Zhang, Cheng Qian, Zeping Li, Xiaoteng Ma, Guanhua Chen*, Heng Ji
- In Proceedings of ACL 2026 (CCF-A, main conference)
VFA: Empoweing Multilingual MLLMs via Vision-Free Adaptation
- Yixia Li, Yaqing Shi, Zhiwen Ruan, Dongdong Zhang, Lingjie Jiang, Shaohan Huang, Yun Chen, Guanhua Chen*, Furu Wei
- In Proceedings of ACL 2026 (CCF-A, main conference)
InstructDiff: Domain-Adaptive Data Selection via Differential Entropy for Efficient LLM Fine-Tuning
- Junyou Su#, He Zhu#,*, Xiao Luo, Liyu Zhang, Hong-Yu Zhou, Yun Chen, Peng Li, Yang Liu, Guanhua Chen*
- In Proceedings of ACL 2026 (CCF-A, main conference)
Modeling LLM Unlearning as an Asymmetric Two-Task Learning Problem
- Zeguan Xiao, Siqing Li, Yong Wang, Xuetao Wei, Jian Yang, Yun Chen*, Guanhua Chen*
- In Proceedings of ACL 2026 (CCF-A, main conference)
Evaluating Memory Capability in Continuous Lifelog Scenario
- Jianjie Zheng, Zhichen Liu, Z. Shen, J. Qu, Guanhua Chen*, Yile Wang, Yang Xu, Yang Liu, Sijie Cheng*
- In Findings of ACL 2026
Toward Automated Robustness Evaluation of Mathematical Reasoning
- Yutao Hou, Zeguan Xiao, Fei Yu, Y. Jiang, S. Ma, Z. Dai, Hailiang Huang*, Yun Chen*, Guanhua Chen*
- In Findings of ACL 2026
FinSafetyBench: Evaluating LLM Safety in Real-World Financial Scenarios
- Yutao Hou, Yihan Jiang, Yuhan Xie, Jian Yang, Liwen Zhang, Hailiang Huang*, Guanhua Chen*, Yun Chen*
- In Findings of ACL 2026
Towards Bridging the Reward-Generation Gap in Direct Alignment Algorithms
- Zeguan Xiao, Yun Chen, Jian Yang, Guanhua Chen*, Ke Tang
- In Findings of ACL 2026
Representation-Guided Parameter-Efficient LLM Unlearning
- Zeguan Xiao, Lang Mo, Yun Chen, Lei Yang, Jiehui Zhao, Lili Yang*, Guanhua Chen*
- In Findings of ACL 2026
Beyond Static Rules: Automated Discovery of Latent Vulnerabilities in Text-to-SQL
- Hanqing Wang, Yongdong Chi, Jian Yang, Lei Yang, Jiehui Zhao, Yun Chen, Guanhua Chen*
- In Findings of ACL 2026
InfoScan: Information-Efficient Visual Scanning via Resource-Adaptive Walks
Anchored Supervised Fine-Tuning
BiasScope: Towards Automated Detection of Bias in LLM-as-a-Judge Evaluation
Compound-QA: A Benchmark for Evaluating LLMs on Compound Questions
Enhancing Uncertainty Estimation in LLMs with Expectation of Aggregated Internal Belief
Beyond the Surface: Enhancing LLM-as-a-Judge Alignment with Human via Internal Representations
G2: Guided Generation for Enhanced Output Diversity in LLMs
Pi-SQL: Enhancing Text-to-SQL with Fine-Grained Guidance from Pivot Programming Languages
ImPart: Importance-Aware Delta-Sparsification for Improved Model Compression and Merging in LLMs
PlanGPT: Enhancing Urban Planning with Tailored Language Model and Efficient Retrieval
- He Zhu, Guanhua Chen*，Wenjia Zhang*.
- In Proceedings of ACL 2025 (industry track) (CCF A, oral, long paper) [pdf]
Fanno: Augmenting High-Quality Instruction Data with Open-Sourced LLMs Only
Tag-Instruct: Controlled Instruction Complexity Enhancement through Structure-based Augmentation
MiLoRA: Harnessing Minor Singular Components for Parameter-Efficient LLM Finetuning
SeqAR: Jailbreak LLMs with Sequential Auto-Generated Characters
Self-DC: When to Reason and When to Act? Self Divide-and-Conquer for Compositional Unknown Questions
- Hongru Wang, Boyang Xue, B. Zhou, T. Zhang, C. Wang, Huimin Wang, Guanhua Chen*, Kam-Fai Wong*
- In Proceedings of NAACL 2025 (CCF B, long paper in main conference) [pdf]
LayAlign: Enhancing Multilingual Reasoning in LLMs via Layer-Wise Adaptive Fusion and Alignment Strategy
SeTAR: Out-of-Distribution Detection with Selective Low-Rank Approximation
Distract Large Language Models for Automatic Jailbreak Attack
PACIT: Unlocking the Power of Examples for Better In-Context Instruction Tuning
mCLIP: Multilingual CLIP via Cross-lingual Transfer
XLM-D: Decorate Cross-lingual Pre-training Model as Non-Autoregressive Neural Machine Translation
Multilingual Sentence Transformer as A Multilingual Word Aligner
Towards Making the Most of Cross-Lingual Transfer for Zero-Shot Neural Machine Translation
Zero-shot Cross-lingual Transfer of Neural Machine Translation with Multilingual Pretrained Encoders
Lexically Constrained Neural Machine Translation with Explicit Alignment Guidance
Lexical-Constraint-Aware Neural Machine Translation via Data Augmentation
Co-Authored Publications
[ACL’26] No More Stale Feedback: Co-Evolving Critics for Open-World Agent Learning
[ACL’26] Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agents
[ACL’26] CAPruner: Conceptual-Adjacent Scene Graph Pruner for Enhancing 3D Spatial Reasoning of LLMs
[ACL’26 Findings] Thinking with Map: Reinforced Parallel Map-Augmented Agent for Geolocalization
[ACL’26 Findings] Can LLMs Hear the Dogwhistle?
[ICLR’26] Tree Search for LLM Agent Reinforcement Learning
[ICLR’26] VisCodex: Unified Multimodal Code Generation via Merging Vision and Coding Models
[ICLR’26] Fair Decision Utility in Human-AI Collaboration: Interpretable Confidence Adjustment for Humans with Cognitive Disparities
[ICLR’26] From Abstract to Contextual: What LLMs Still Cannot Do in Mathematics
[ICLR’26] Synthesizing Multimodal Verifiable Game Data to Boost VLMs’ General Reasoning
[AAAI’26] ConInstruct: Evaluating Large Language Models on Conflict Detection and Resolution in Instructions
[NeurIPS’25] Alleviating Hallucinations in Large Language Models through Multi-Model Contrastive Decoding and Dynamic Hallucination Detection
[EMNLP’25] PlanGPT-VL: Enhancing Urban Planning with Domain-Specific Vision-Language Models
[RAL’25] A Joint Learning of Force Feedback of Robotic Manipulation and Textual Cues for Granular Materials Classification
[ACL’25] LLMs Trust Humans More, That’s a Problem! Unveiling and Mitigating the Authority Bias in Retrieval-Augmented Generation
[ACL’25 Findings] The Elephant in the Room: Exploring the Role of Neutral Words in Language Model Group-Agnostic Debiasing
[RAL’25] Understanding Particles from Video: Property Estimation of Granular Materials via Visuo-Haptic Learning
[EMNLP’23 Findings] StyleBART: Decorate Pretrained Model with Style Adapters for Unsupervised Stylistic Headline Generation
[ECAI’23] Evaluating Explanation Methods for Vision-and-Language Navigation
[EMNLP’20] Accurate Word Alignment Induction from Neural Machine Translation
Working Papers
（all are corresponding authored papers）
UniRRM: Unified Reasoning Reward Models Across Languages and Evaluation Paradigms
GraviScan: Information Gravitation Modeling for Efficient Visual Scanning
Anchored Policy Optimization: Mitigating Exploration Collapse via Support-Constrained Rectification
Rethinking Robust LLM Unlearning Against Relearning Attacks: The Minor Components in Representations Matter
Safety-Preserving Adaptation via Fine-Tuning Transfer for Large Language Models
Towards Fair And Comprehensive Evaluation Of Routers In Collaborative LLM Systems
StatABench: Dataset and Framework for Evaluating Statistical Analysis Capabilities of LLMs
BAM-MT: Batch-Adaptive Multi-Objective Reinforcement Learning for Medical Machine Translation
AlignDiff: Exploiting Model-Intrinsic Information for Better Preference Data Selection
Enhancing Large Language Model Reasoning via Selective Critical Token Fine-Tuning
Unveiling Over-Memorization in Finetuning LLMs for Reasoning Tasks
Enhancing Delta Compression in LLMs via SVD-based Quantization Error Minimization
Teaching
- Undergraduate course STA323: Big Data Analysis Software and Application (Hadoop or Spark)
- Graduate course STA5007: Advanced Natural Language Processing
Group Members
- PhD candidates: Peng Lai(2026), Yunpeng Sun(SLAI,2025), Zhiwen Ruan(2024), Yixia Li(2023)
- Master students: Xuanzhe Xu(2026), Qianyu Yang(2026), Qi Wang(2025), Youxin Zhu(2025), Xiaodong Lao(2024), Jianjie Zheng(2024)
- Visiting students: Zeguan Xiao, He Zhu, Yifeng Wu, Boya Xiong, Yutao Hou, Yongdong Chi, Wanxing Wu, Tianyi Wang, Lecheng Yan
(Last updated on Apr. 7, 2026)