Jinghui Zhang, Lang Gao, Zongfang Liu, Ruihong Zeng, Zirui Song, Rui Yan, Kentaro Inui, Xiuying Chen

NeurIPS 2026

CopyGen is a lightweight decoding-time plugin that makes LLMs reproduce source spans more exactly and faster, improving average exact match by 35% relative without finetuning.

Jinghui Zhang, Kaiyang Wan, Longwei Xu, Ao Li, Zongfang Liu, Xiuying Chen

ACM MM 2025

Proposed a personalized response prediction framework with PAC-LoRA, enabling user-specific comment generation while modeling crowd-level sentiment trends.

Jinghui Zhang, Lang Gao, Ao Li, Mingzhe Li, Ruihong Zeng, Zirui Song, Kentaro Inui, Xiuying Chen

EMNLP 2026 Findings

We reframe different writing patterns into a unified interpretable style space, enabling personalized author-aware text generation framework via steering outputs toward target authorial characteristics.

Lang Gao^*, Jinghui Zhang^*, Wei Liu, Fengxian Ji, Chenxi Wang, Zirui Song, Akash Ghosh, Youssef Mohamed, Preslav Nakov, Xiuying Chen

ICML 2026     [Invited Talk @ Cohere]

Introduced CRH, a cylindrical geometric model to explain instability in activation steering, and empirically verified its sample-specific structure across models, concepts, and steering methods.

Ao Li^*, Jinghui Zhang^*, Luyu Li, Yuxiang Duan, Lang Gao, Mingcai Chen, Weijun Qin, Shaopeng Li, Fengxian Ji, Ning Liu, Lizhen Cui, Xiuying Chen, Yuntao Du

ACM MM 2026

M3MAD-Bench is a standardized multimodal benchmark for assessing multi-agent debate methods across domains, modalities, and performance-cost trade-offs.

Ao Li, Yuxiang Duan, Jinghui Zhang, Congbo Ma, Yutong Xie, Gustavo Carneiro, Mohammad Yaqub, Hu Wang

CVPR 2026

TransPrune is a training-free token pruning method that combines Token Transition Variation with instruction-guided attention to reduce LVLM inference cost by over 50% with comparable performance.