Papers by Jinbao Li
Reinforcement Learning on Pre-Training Data (2026.acl-long)
Copied to clipboard
Siheng Li, Kejiao Li, Zenan Xu, Guanhua Huang, Kun Li, Haoyuan Wu, null Wujiajia, Zihao Zheng, Chenchen Zhang, Kun Shi, Xue Gong, Qi Yi, Ruibin Xiong, Tingqiang Xu, Yuhao Jiang, Jianfeng Yan, Yuyuan Zeng, Guanghui Xu, Jinbao Xue, Zhijiang xu, Zheng Fang, Shuai LI, Qibin Liu, Xiaoxue Li, Zhuoyu Li, Yangyu Tao, Fei Gao, Cheng Jiang, Bochao Wang, Kai Liu, Jianchen Zhu, Wai Lam, Bo Zhou, Di Wang
| Challenge: | Recent progress in large language models is driven by scaling of training compute through pre-training with nexttoken prediction (NTP) or post-training (RL) Pre-training using NTP enables models to acquire extensive knowledge and skills from general data, but it suffers from data inefficiency and catastrophic forgetting in continual learning settings. |
| Approach: | They propose to scale training compute through pre-training with next-token prediction (NTP) or post-training by scaling reinforcement learning (RL) to improve learning from general data. |
| Outcome: | Experiments on multiple benchmarks and models show that the proposed approach improves continual pre-training and provides a strong foundation for post-training on Qwen3-8B-Base. |
“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. |
TWIST: Text-encoder Weight-editing for Inserting Secret Trojans in Text-to-Image Models (2025.acl-long)
Copied to clipboard
| Challenge: | Existing Trojan attacks require extensive training data and poor generalization, limiting effectiveness and scalability. |
| Approach: | They propose a method for embedding Trojans into plugins using a single edit layer . they find that the method reduces modified parameters by 8-fold and cuts injection time to 25 seconds . |
| Outcome: | The proposed method achieves an average attack success rate of 91%, a 78% improvement over the state-of-the-art (SOTA) method. |
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. |
PerMemSafe: Benchmarking Implicit Personalized Safety of Long Horizon Self-Evolving Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing self-evolving agents have a low safety rate in long-horizon interactions . however, this reliance on context-independent safety evaluations is insufficient . |
| Approach: | They propose a framework that explicitly models personalized risk inference and memory evolution. |
| Outcome: | The proposed framework improves implicit personalized safety by 23.8% over prior frameworks while maintaining helpfulness in long-horizon interactions. |