Research

My current main research interest is on building efficient and reliable LLM/agent systems, which involves research at two levels:

• The Model Level: Enhancing long-context capabilities, Understanding LM mechanisms, …
  (Gist-Compression, Gist-Edit, Attention-Entropy-Analysis, …)
• The System Level: Better search methods, Better system designs, …
  (Web-Rollback, Cognitive Kernel-Pro, …)

Papers

Here is a list of my selected papers:

[Google Scholar] [Semantic Scholar] [dblp]

• WebRollback: Enhancing Web Agents with Explicit Rollback Mechanisms. [arxiv]
  Z. Zhang, T. Fang, K. Ma, W. Yu, H. Zhang, H. Mi, D. Yu. In Proceedings of EACL, 2026.

• Cognitive Kernel-Pro: A Framework for Deep Research Agents and Agent Foundation Models Training. [arxiv] [system]
  T. Fang, Z. Zhang, X. Wang, R. Wang, C. Qin, Y. Wan, J. Ma, C. Zhang, J. Chen, X. Li, H. Zhang, H. Mi, D. Yu. Technical Report, 2025.

• InComeS: Integrating Compression and Selection Mechanisms into LLMs for Efficient Model Editing. [arxiv]
  S. Li, Z. Zhang^†, Y. Deng, C. Deng, T. Fang, H. Zhang, H. Mi, D. Yu, W. Lam^†. Preprint, 2025.

• UniGist: Towards General and Hardware-aligned Sequence-level Long Context Compression. [arxiv]
  C. Deng, Z. Zhang^†, K. Mao, S. Li, T. Fang, H. Zhang, H. Mi, D. Yu, and Z. Dou^†. In Proceedings of NeurIPS, 2025.

• UNCLE: Uncertainty Expressions in Long-Form Generation. [arxiv] [paper]
  R. Yang*, C. Zhang*, Z. Zhang^†, X. Huang, D. Yu, N. Collier^†, and D. Yang^†. In Proceedings of EMNLP, 2025.

• Attention Entropy is a Key Factor: An Analysis of Parallel Context Encoding with Full-attention-based Pre-trained Language Models. [arxiv] [paper]
  Z. Zhang, Y. Wang, X. Huang, T. Fang, H. Zhang, C. Deng, S. Li, and D. Yu. In Proceedings of ACL, 2025.

• A Silver Bullet or a Compromise for Full Attention? A Comprehensive Study of Gist Token-based Context Compression. [arxiv] [paper]
  C. Deng, Z. Zhang^†, K. Mao, S. Li, X. Huang, D. Yu, and Z. Dou^†. In Proceedings of ACL, 2025.

• LoGU: Long-form Generation with Uncertainty Expressions. [arxiv] [paper]
  R. Yang*, C. Zhang*, Z. Zhang^†, X. Huang, S. Yang, N. Collier, D. Yu, and Deqing Yang^†. In Proceedings of ACL, 2025.

• Atomic Calibration of LLMs in Long-Form Generations. [arxiv] [paper]
  C. Zhang, R. Yang, Z. Zhang^†, X. Huang, S. Yang, D. Yu, and N. Collier^†. In Proceedings of AACL-IJCNLP-Findings, 2025.

• Data-efficient Active Learning for Structured Prediction with Partial Annotation and Self-Training. [paper]
  Z. Zhang, E. Strubell, and E. Hovy. In Proceedings of EMNLP-Findings, 2023.

• On the Interactions of Structural Constraints and Data Resources for Structured Prediction. [paper]
  Z. Zhang, E. Strubell, and E. Hovy. In Proceedings of SustaiNLP, 2023.

• Towards More Efficient Insertion Transformer with Fractional Positional Encoding. [paper] [code]
  Z. Zhang, Y. Zhang, and B. Dolan. In Proceedings of EACL, 2023.

• A Survey of Active Learning for Natural Language Processing. [paper] [table]
  Z. Zhang, E. Strubell, and E. Hovy. In Proceedings of EMNLP, 2022.

• Transfer Learning from Semantic Role Labeling to Event Argument Extraction with Template-based Slot Querying. [paper] [code]
  Z. Zhang, E. Strubell, and E. Hovy. In Proceedings of EMNLP, 2022.

• On the Benefit of Syntactic Supervision for Cross-lingual Transfer in Semantic Role Labeling. [paper] [code]
  Z. Zhang, E. Strubell, and E. Hovy. In Proceedings of EMNLP, 2021.

• Comparing Span Extraction Methods for Semantic Role Labeling. [paper] [code]
  Z. Zhang, E. Strubell, and E. Hovy. In Proceedings of SPNLP, 2021.

• An Empirical Exploration of Local Ordering Pre-training for Structured Prediction. [paper] [code]
  Z. Zhang, X. Kong, L. Levin, and E. Hovy. In EMNLP-Findings, 2020.

• Incorporating a Local Translation Mechanism into Non-autoregressive Translation. [paper]
  X. Kong*, Z. Zhang*, and E. Hovy. In Proceedings of EMNLP, 2020.

• A Two-Step Approach for Implicit Event Argument Detection. [paper] [code]
  Z. Zhang, X. Kong, Z. Liu, X. Ma and E. Hovy. In Proceedings of ACL, 2020.

• An Empirical Investigation of Structured Output Modeling for Graph-based Neural Dependency Parsing. [paper] [code]
  Z. Zhang, X. Ma and E. Hovy. In Proceedings of ACL, 2019.

• On Difficulties of Cross-Lingual Transfer with Order Differences: A Case Study on Dependency Parsing. [paper] [code1] [code2]
  W. Ahmad*, Z. Zhang*, X. Ma, E. Hovy, K. Chang and N. Peng. In Proceedings of NAACL, 2019.

• Exploring Recombination for Efficient Decoding of Neural Machine Translation. [paper] [code]
  Z. Zhang, R.Wang, M. Utiyama, E. Sumita and H. Zhao. In Proceedings of EMNLP, 2018.

• Probabilistic Graph-based Dependency Parsing with Convolutional Neural Network. [paper] [code]
  Z. Zhang, H. Zhao and L. Qin. In Proceedings of ACL, 2016.

• High-order Graph-based Neural Dependency Parsing. [paper] [code]
  Z. Zhang and H. Zhao. In Proceedings of PACLIC-29, 2015.