About

Dongming Jiang is a Ph.D. Candidate in Computer Science at The University of Texas at Dallas, advised by Dr. Bingzhe Li. His research focuses on agentic memory architectures, long-horizon reasoning, retrieval systems, and scalable machine-learning infrastructure.

He builds reliable end-to-end AI systems spanning memory, retrieval, evaluation, and systems optimization.

Agentic memory · LLM agents · Long-horizon reasoning · Retrieval systems · ML infrastructure

Featured Research

Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents

Preprint

A System-One-controlled agentic memory architecture that moves frequent structured memory decisions out of the autoregressive generation loop while reserving System-Two language models for deeper reasoning and answer synthesis.

Reported experimental results on LoCoMo with GPT-4o-mini: 0.777 overall score, 158 s memory construction, and 0.93 s average query latency.

• Paper (arXiv)
• Hugging Face Papers
• Project Page
• Code

MAGMA: A Multi-Graph-based Agentic Memory Architecture for AI Agents

ACL 2026 Main Conference

A multi-graph memory architecture for AI agents that organizes and retrieves long-term information through complementary relational views.

• Paper (ACL Anthology)
• arXiv
• Hugging Face Papers
• Code

HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution

Preprint · Under Review

An agentic-memory method that uses reinforcement learning to evolve weighted memory graphs dynamically.

• Paper (arXiv)
• Hugging Face Papers
• Code

Adaptive Debiasing Tsallis Entropy for Test-Time Adaptation

ICLR 2026

A test-time adaptation method for vision-language models that replaces Shannon entropy with an adaptive Tsallis entropy, using a class-specific non-extensive parameter derived from the label bias estimated over incoming test instances.

• Paper (OpenReview)
• arXiv
• Code

News

Selected Publications — see the full list

Dongming Jiang in bold. ^* equal (co-first) contribution · ^† corresponding author.

• ACL ’26

  MAGMA: A Multi-Graph-based Agentic Memory Architecture for AI Agents

  Dongming Jiang^*, Yi Li, Guanpeng Li, Bingzhe Li

  Annual Meeting of the Association for Computational Linguistics (Main Conference), 2026

  • Paper
  • arXiv
  • Code
• Preprint

  Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents

  Dongming Jiang, Yi Li, Bingzhe Li

  arXiv preprint, 2026

  • Paper
  • Project Page
  • Code
• ICLR ’26

  Adaptive Debiasing Tsallis Entropy for Test-Time Adaptation

  Xiangyu Wu^*, Dongming Jiang^*, F. Yu, Y. Tian, J. Tang, Q.-G. Chen, Y. Yang, J. Lu

  International Conference on Learning Representations, 2026

  • Paper
  • Code
• Under ReviewPreprint

  Anatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System Limitations

  Dongming Jiang^*†, Yi Li, Songtao Wei, Jinxin Yang, Ayushi Kishore, Alysa Zhao, Dingyi Kang, Xu Hu, Feng Chen, et al.

  arXiv preprint, 2026

  • Paper
  • Paper Collection
• Under ReviewPreprint

  HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution

  Dongming Jiang^*, Yi Li, Guanpeng Li, Qiannan Li, Bingzhe Li

  arXiv preprint, 2026

  • Paper
  • Code

Education

Academic Service

Reviewer / Program Committee

• NeurIPS
• ICLR
• ICML
• CVPR
• TMLR
• ICCAD
• OSDI (Artifact Evaluation)
• PEARC
• MMM
• IEEE Security & Privacy

Technical skills

Python · C++ · CUDA · PyTorch · TensorFlow · Hugging Face · Linux · Docker · SLURM · MPI