Papers by Ziwen Wang
AMO-Bench: Large Language Models Still Struggle in High School Math Competitions (2026.findings-acl)
Copied to clipboard
Junlin Liu, Shengnan An, Shuang Zhou, Dan Ma, Yehao Lin, Xinxuan Lv, Xuanlin Wang, Xiaoyu Li, Ziwen Wang, Xuezhi Cao, Xunliang Cai
| Challenge: | Existing benchmarks for mathematical reasoning are becoming less effective due to performance saturation. |
| Approach: | They propose to use a mathematical reasoning benchmark with Olympiad difficulty to evaluate top-tier LLMs. |
| Outcome: | The proposed benchmarks are cross-validated by experts to meet IMO difficulty standards and entirely original problems to prevent performance leakages from data memorization. |
Detoxifying Large Language Models via Knowledge Editing (2024.acl-long)
Copied to clipboard
Mengru Wang, Ningyu Zhang, Ziwen Xu, Zekun Xi, Shumin Deng, Yunzhi Yao, Qishen Zhang, Linyi Yang, Jindong Wang, Huajun Chen
| Challenge: | Existing methods to detoxify Large Language Models (LLMs) are limiting, but knowledge editing can be effective. |
| Approach: | They propose a baseline method to detoxify Large Language Models (LLMs) they propose supervised fine-tuning and reinforcement learning from human feedback (RLHF) |
| Outcome: | The proposed method reduces toxicity of large language models with one instance of tuning . it reduces the toxicity, while minimizing the toxins, the authors show . |
Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atoms (2025.acl-long)
Copied to clipboard
| Challenge: | Recent research has explored the use of sparse autoencoders (SAE) to disentangle knowledge in high-dimensional spaces for steering. |
| Approach: | They propose a method that isolates and manipulates disentangled knowledge components to enhance safety by using sparse autoencoders to disentangle knowledge in high-dimensional spaces for steering. |
| Outcome: | The proposed method is able to isolate and manipulate disentangled knowledge components to enhance safety in large reasoning models. |
FACT: Examining the Effectiveness of Iterative Context Rewriting for Multi-fact Retrieval (2025.findings-naacl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are proficient at retrieving single facts from extended contexts, but struggle with tasks requiring simultaneous retrieval of multiple facts. |
| Approach: | They propose a method that refines context through successive rounds of rewriting to address this problem by finding all Crucial Texts (FACT) |
| Outcome: | The proposed method improves multi-fact retrieval performance across tasks, though improvements are less notable in general-purpose QA scenarios. |
Automating Steering for Safe Multimodal Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in Multimodal Large Language Models (MLLMs) have unlocked powerful cross-modal reasoning abilities, but also raised new safety concerns, especially when faced with adversarial multimodal inputs. |
| Approach: | They propose a modular and adaptive inference-time intervention technology, AutoSteer, that integrates a safety awareness score, an adaptive safety prober, and a lightweight Refusal Head to modulate generation when safety risks are detected. |
| Outcome: | Experiments on LLaVA-OV and Chameleon show that AutoSteer significantly reduces the Attack Success Rate (ASR) for textual, visual, and cross-modal threats while maintaining general abilities. |
EasyEdit: An Easy-to-use Knowledge Editing Framework for Large Language Models (2024.acl-demos)
Copied to clipboard
Peng Wang, Ningyu Zhang, Bozhong Tian, Zekun Xi, Yunzhi Yao, Ziwen Xu, Mengru Wang, Shengyu Mao, Xiaohan Wang, Siyuan Cheng, Kangwei Liu, Yuansheng Ni, Guozhou Zheng, Huajun Chen
| Challenge: | Large Language Models (LLMs) suffer from knowledge cutoff or fallacy issues, which means they are unaware of unseen events or generate text with incorrect facts owing to outdated/noisy data. |
| Approach: | They propose an easy-to-use knowledge editing framework for Large Language Models that allows users to easily edit updated knowledge and adjust undesired behavior while minimizing the impact on unrelated inputs. |
| Outcome: | The proposed framework surpasses traditional fine-tuning in terms of reliability and generalization. |
Knowledge Mechanisms in Large Language Models: A Survey and Perspective (2024.findings-emnlp)
Copied to clipboard
Mengru Wang, Yunzhi Yao, Ziwen Xu, Shuofei Qiao, Shumin Deng, Peng Wang, Xiang Chen, Jia-Chen Gu, Yong Jiang, Pengjun Xie, Fei Huang, Huajun Chen, Ningyu Zhang
| Challenge: | Using large language models, we can understand knowledge mechanisms in LLMs for learning, storage, utilization, and evolution. |
| Approach: | They propose to analyze knowledge mechanisms in Large Language Models (LLMs) they examine utilization, evolution, and the potential dark knowledge (hypothesis) they hope to help understand knowledge in LLMs and provide insights for future research . |
| Outcome: | The proposed model can be used to analyze the evolution of parametric knowledge in LLMs. |
EasyEdit2: An Easy-to-use Steering Framework for Editing Large Language Models (2025.emnlp-demos)
Copied to clipboard
Ziwen Xu, Shuxun Wang, Kewei Xu, Haoming Xu, Mengru Wang, Xinle Deng, Yunzhi Yao, Guozhou Zheng, Huajun Chen, Ningyu Zhang
| Challenge: | Large Language Models (LLMs) have demonstrated extraordinary capabilities, however, they may still generate unreliable or unsafe outputs. |
| Approach: | They propose a framework that allows plug-and-play adjustability for controlling Large Language Model (LLM) behaviors. |
| Outcome: | The framework is designed to enable plug-and-play adjustability for controlling Large Language Model (LLM) behaviors. |
Why Steering Works: Toward a Unified View of Language Model Parameter Dynamics (2026.acl-long)
Copied to clipboard
Ziwen Xu, Chenyan WU, Hengyu Sun, Haiwen Hong, Mengru Wang, Yunzhi Yao, Longtao Huang, Hui Xue, Shumin Deng, Zhixuan Chu, Huajun Chen, Ningyu Zhang
| Challenge: | Methods for controlling large language models (LLMs) are often studied in isolation, obscuring connections and making comparison difficult. |
| Approach: | They propose a preference-utility analysis that separates control effects into preference and utility, and measures both on a shared log-odds scale using polarity-paired contrastive examples. |
| Outcome: | The proposed approach improves preference while preserving utility. |