Papers by Wenhai Wang
Watch Out Your Industrial Copilots: Stealthy Backdoor Attack Against LLM-Based PLC Code Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are being used to generate PLC code from natural language. |
| Approach: | They propose a stealthy backdoor attack framework targeting LLM-based PLC code generation . they incorporate six malicious logic injection patterns and a pipeline to refine stealthiness . |
| Outcome: | The proposed framework achieves 82.92% success rate while remaining stealthy . it bypasses quality validation and is difficult to detect . |
CP-BCS: Binary Code Summarization Guided by Control Flow Graph and Pseudo Code (2023.emnlp-main)
Copied to clipboard
| Challenge: | Current work on understanding assembly code is oriented towards generating function names, which involve numerous abbreviations that make them confusing. |
| Approach: | They propose a control flow graph and pseudo code guided binary code summarization framework to learn the comprehensive binary function execution behavior and logic semantics. |
| Outcome: | The proposed framework improves the efficiency of reverse engineering on 3 different binary optimization levels for 3 different computer architectures. |
LLM-VA: Resolving the Jailbreak-Overrefusal Trade-off via Vector Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Existing vector steering methods adjust the magnitude of answer vectors, but this creates a fundamental trade-off—reducing jailbreak increases over-refusal. |
| Approach: | They propose a method which aligns va with vb through closed-form weight updates, making the model’s willingness to respond causally dependent on its safety assessment. |
| Outcome: | Experiments on 12 LLMs show that the proposed method achieves 11.45% higher F1 than the best baseline while preserving 95.92% utility. |
Sticking to the Mean: Detecting Sticky Tokens in Text Embedding Models (2025.acl-long)
Copied to clipboard
| Challenge: | Sticky tokens, when repeatedly inserted into sentences, pull sentence similarity toward a certain value, disrupting the normal distribution of embedding distances and degrading downstream performance. |
| Approach: | They propose a method to detect “sticky tokens” by sentence and token filtering and apply it to 40 checkpoints across 14 model families. |
| Outcome: | The proposed method detects 868 sticky tokens across 14 models and shows that their presence does not correlate with model size or vocabulary size. |
Selective Knowledge Distillation: Fusing LLM Semantic Strengths with DNN Efficiency for Binary Code Similarity Detection (2026.acl-long)
Copied to clipboard
| Challenge: | BinSKD is a binary code similarity detection technique that can be used in bug detection, patch analysis, and malware detection. |
| Approach: | They propose to leverage an LLM-based BCSD method as the teacher model and transfer its knowledge of high-level program semantics to various DNN-based student models. |
| Outcome: | The proposed method yields Recall@1 improvements of 14.5%–91.2% for DNN-based BCSD methods and enables HermesSim to match the teacher’s performance with orders-of-magnitude efficiency. |
SDAR: A Synergistic Diffusion-AutoRegression Paradigm for Scalable Sequence Generation (2026.findings-acl)
Copied to clipboard
Shuang Cheng, Yihan Bian, Dawei Liu, Yuhua Jiang, Yihao Liu, Linfeng Zhang, Qian Yao, Zhongbo Tian, Wenhai Wang, Qipeng Guo, Kai Chen, Biqing Qi, Bowen Zhou
| Challenge: | Autoregressive (AR) language models are a dominant paradigm in the field of parallelism and non-causal modeling. |
| Approach: | They propose a blockwise discrete diffusion model that preserves AR-compatible serving while enabling parallel intra-block generation. |
| Outcome: | The proposed model achieves theoretical speedups over 5 and wall-clock speedup of 2.3 on H200 GPUs in latency-critical regimes. |
Tram: A Token-level Retrieval-augmented Mechanism for Source Code Summarization (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods to generate source code summaries are coarse-grained and noise-filled . however, they do not capture contextual code semantics and are often outdated in continuous software iteration. |
| Approach: | They propose a fine-grained Token-level retrieval-augmented mechanism on the decoder side to enhance performance of neural models. |
| Outcome: | The proposed method produces more low-frequency tokens and is interpretable. |
OmniAlign-V: Towards Enhanced Alignment of MLLMs with Human Preference (2025.acl-long)
Copied to clipboard
Xiangyu Zhao, Shengyuan Ding, Zicheng Zhang, Haian Huang, Maosongcao Maosongcao, Jiaqi Wang, Weiyun Wang, Xinyu Fang, Wenhai Wang, Guangtao Zhai, Hua Yang, Haodong Duan, Kai Chen
| Challenge: | Existing open-source multi-modal large language models (MLLMs) focus on enhancing foundational capabilities, leaving a significant gap in human preference alignment. |
| Approach: | They propose a dataset of 200K high-quality training samples featuring diverse images, complex questions, and varied response formats to improve MLLMs’ alignment with human preferences. |
| Outcome: | The proposed dataset of 200K high-quality training samples improves human preference alignment while maintaining or enhancing performance on standard VQA benchmarks. |