Papers by Zhenting Wang
NOTABLE: Transferable Backdoor Attacks Against Prompt-based NLP Models (2023.acl-long)
Copied to clipboard
| Challenge: | Existing backdoor attacks against prompt-based learning involve injecting back doors into embedding layers or word embedders. |
| Approach: | They propose a backdoor attack against prompt-based learning that injects backdoors into embedding layers or word embeddable vectors. |
| Outcome: | The proposed backdoor attack outperforms two state-of-the-art models on six NLP tasks and three prompting strategies. |
An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing jailbreaking methods generate harmful and unethical content when subjected to jailbreaking attacks. |
| Approach: | They propose a black-box jailbreaking method with optimizable suffixes that translate jailbreaking objectives into natural language instructions. |
| Outcome: | The proposed method outperforms existing methods by 2.4 times in the ASR of three open-source LLMs and GPT-3.5-Turbo. |
Reasoning over Precedents Alongside Statutes: Case-Augmented Deliberative Alignment for LLM Safety (2026.acl-long)
Copied to clipboard
Can Jin, Rui Wu, Tong Che, Qixin Zhang, Hongwu Peng, Jiahui Zhao, Zhenting Wang, Wenqi Wei, Ligong Han, Zhao Zhang, Yuan Cao, Ruixiang Tang, Dimitris N. Metaxas
| Challenge: | OpenAI introduces deliberative alignment (DA) to enhance safety of its o-series models, but effectiveness of this approach in open-source LLMs is understudied. |
| Approach: | They propose a case-augmented deliberative alignment method for large language models . they propose to use reinforcement learning on self-generated safety reasoning chains . |
| Outcome: | The proposed method avoids narrowly enumerated rules and allows broader adaptability. |
Exploring Concept Depth: How Large Language Models Acquire Knowledge and Concept at Different Layers? (2025.coling-main)
Copied to clipboard
Mingyu Jin, Qinkai Yu, Jingyuan Huang, Qingcheng Zeng, Zhenting Wang, Wenyue Hua, Haiyan Zhao, Kai Mei, Yanda Meng, Kaize Ding, Fan Yang, Mengnan Du, Yongfeng Zhang
| Challenge: | Large language models have shown remarkable performances across a wide range of tasks, but mechanisms by which they encode tasks of varying complexity remain poorly understood. |
| Approach: | They propose to explore the possibility that LLMs process concepts in different layers . they propose to categorize concepts based on their level of abstraction . |
| Outcome: | The proposed model can process complex concepts in shallow layers, the authors show . the proposed model could be used to prob complex tasks in shallow ones . |
Data-centric NLP Backdoor Defense from the Lens of Memorization (2025.findings-naacl)
Copied to clipboard
| Challenge: | Backdoor attacks pose a severe threat to the trustworthiness of DNN-based language models. |
| Approach: | They propose a data-centric defense that extends memorization definitions to fine-grained sentences . they find that duplicated sentence elements are necessary for successful backdoor attacks . |
| Outcome: | The proposed defense outperforms state-of-the-art defenses against backdoor attacks. |
Token-Budget-Aware LLM Reasoning (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to enhance reasoning capabilities of large language models incur significant overhead in token usage, leading to increased costs. |
| Approach: | They propose a token-budget-aware LLM reasoning framework that adjusts the number of reasoning tokens based on the reasoning complexity of each problem. |
| Outcome: | The proposed method reduces token costs in CoT reasoning with only a slight performance reduction. |
ADO: Automatic Data Optimization for Inputs in LLM Prompts (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent research has focused on refining instruction components and augmenting input data with in-context examples, but this study explores the potential benefits of optimizing the input data itself. |
| Approach: | They propose a content engineering and structural reformulation strategy to optimize input data within prompts to improve performance of Large Language Models. |
| Outcome: | The proposed approach improves performance of Large Language Models (LLMs) in various tasks, offering a promising avenue for future research in prompt engineering. |
Debiasing LLMs by Masking Unfairness-Driving Attention Heads (2026.findings-acl)
Copied to clipboard
Tingxu Han, Wei Song, Ziqi Ding, Ziming Li, Chunrong Fang, Yuekang Li, Dongfang Liu, Zhenyu Chen, Zhenting Wang
| Challenge: | Existing work probes when biased outputs appear, but gives little insight into the mechanisms that generate them, leaving existing mitigations largely fragile. |
| Approach: | They propose a lightweight debiasing framework that detects bias heads and selectively masks only those heads that activate under DA and CoT. |
| Outcome: | The proposed framework reduces unfairness by 391.9%- 534.5% in both one- and two-turn dialogues. |
EmojiPrompt: Generative Prompt Obfuscation for Privacy-Preserving Communication with Cloud-based LLMs (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) have substantially expanded their applicability across diverse fields, such as personalized recommendations, health report analysis, and financial decision-making. |
| Approach: | They propose a generative transformation paradigm that obfuscates user data with linguistic and non-linguistic elements before submitting it to cloud-based LLMs. |
| Outcome: | The proposed paradigm obfuscates user private data while maintaining performance compared to the unobflated version. |