Papers by Changjiang Li
VideoEraser: Concept Erasure in Text-to-Video Diffusion Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Experimental results show that VideoEraser outperforms prior methods regarding efficacy, integrity, fidelity, robustness, and generalizability. |
| Approach: | They propose a training-free framework that prevents T2V diffusion models from generating videos with undesirable concepts even when explicitly prompted with those concepts. |
| Outcome: | The proposed framework outperforms existing methods in erasure, celebrity erasion, and explicit content erasing tasks. |
“I See What You Did There”: Can Large Vision-Language Models Understand Multimodal Puns? (2026.acl-long)
Copied to clipboard
Naen Xu, Jiayi Sheng, Changjiang Li, Chunyi Zhou, Yuyuan Li, Tianyu Du, Jun Wang, Zhihui Fu, Jinbao Li, Shouling Ji
| Challenge: | Puns are a common form of rhetorical wordplay that exploits polysemy and phonetic similarity to create humor. |
| Approach: | They propose a multimodal pun generation pipeline and a model to evaluate their understanding of puns. |
| Outcome: | The proposed benchmark improves the understanding of multimodal puns by 16.5% in the F1 test. |
Measuring Meaning Composition in the Human Brain with Composition Scores from Large Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | Existing computational metric to quantify extent of meaning composition is lacking . despite extensive neurolinguistic research, understanding how meaning is constructed is difficult . |
| Approach: | They propose a computational metric to quantify the degree of meaning composition . they use key-value memory interpretation of transformer feed-forward network blocks to study meaning composition. |
| Outcome: | Experimental results show that the metric correlates with brain clusters associated with word frequency, structural processing, and general sensitivity to words, suggesting multifaceted nature of meaning composition during human sentence comprehension. |
ACIArena: Toward Unified Evaluation for Agent Cascading Injection (2026.acl-long)
Copied to clipboard
Hengyu An, Minxi Li, Jinghuai Zhang, Naen Xu, Chunyi Zhou, Changjiang Li, Xiaogang Xu, Tianyu Du, Shouling Ji
| Challenge: | Existing studies consider only limited attack strategies and simplified MAS settings, limiting their generalizability and comprehensive evaluation. |
| Approach: | They propose a framework to evaluate the robustness of Multi-Agent Systems (MAS) they propose unified evaluation suites spanning attack surfaces and attack objectives . |
| Outcome: | ACIArena provides a benchmark of 1,356 test cases for evaluating MAS robustness . it covers six widely used MAS implementations and provides measurable results . |
Multilingual Pretraining and Instruction Tuning Improve Cross-Lingual Knowledge Alignment, But Only Shallowly (2024.naacl-long)
Copied to clipboard
| Challenge: | Current large language models show imbalance abilities in different languages . authors propose two approaches to improve cross-lingual knowledge alignment . |
| Approach: | They propose a framework to assess cross-lingual knowledge alignment of large language models . they propose multilingual pretraining and multilingual instruction tuning to address this problem . |
| Outcome: | The proposed framework assesses the cross-lingual knowledge alignment of LLMs in performance, consistency and conductivity levels. |
Roles of Scaling and Instruction Tuning in Language Perception: Model vs. Human Attention (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent large language models (LLMs) have shown strong abilities to understand natural language, but how these factors affect the models’ language perception is unclear. |
| Approach: | They compare the self-attention of several existing large language models in different sizes to assess the effect of scaling and instruction tuning on language perception. |
| Outcome: | The proposed models are closer to non-native speakers than native speakers in attention, suggesting a sub-optimal language perception of all models. |
Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors (2026.acl-long)
Copied to clipboard
Rui Yin, Tianxu Han, Naen Xu, Changjiang Li, Ping He, Chunyi Zhou, Jun Wang, Zhihui Fu, Tianyu Du, Jinbao Li, Shouling Ji
| Challenge: | Existing methods to inject safety-aligned large language models rely on token-level mappings, which do not guarantee sustained harmful output. |
| Approach: | They propose a method that directly modifies model weights to map a trigger to an attacker-specified response. |
| Outcome: | The proposed method achieves high triggered attack success while maintaining non-triggered safety and general utility. |