Welcome to the official website of the ICDM 2026 workshop on LLM for Tensor Analysis & Tensor for LLM Advancement.
LLMTENA 2026 focuses on the emerging synergy between Large Language Models (LLMs) and tensor analysis. The central theme is a bidirectional research exchange: using LLMs to advance tensor-based data analysis, and using tensor methods to improve the efficiency, interpretability, robustness, and scalability of LLMs.
1.Bridge two research communities that have largely developed in parallel: tensor methods and LLMs.
2.Explore how LLMs can enhance tensor analysis, including tensor completion, decomposition, denoising, visualization, and structured data reasoning.
3.Investigate how tensor methods can improve LLMs, especially through compression, acceleration, interpretability, multi-modal fusion, and efficient fine-tuning.
4.Promote interdisciplinary research combining mathematical foundations, data mining, machine learning, and modern AI.
5.Create a platform for new collaborations and high-quality research on tensor–LLM integration within the ICDM community.
The workshop is intended for researchers, students, and practitioners working in: tensor decomposition, tensor networks, and multilinear algebra, large language models and foundation models, data mining and machine learning, multi-modal learning and structured data reasoning, model compression, acceleration, and efficient AI, knowledge graphs, recommendation systems, neuroscience, spatio-temporal data, and multi-agent systems. Overall, the target audience includes both tensor-method researchers interested in LLMs and LLM researchers seeking mathematically grounded tools for efficiency, interpretability, and structured data analysis.
Looking forward to seeing you at Shenyang.
University of Edinburgh, UK
Guangdong University of Technology, China
{{Affiliation}}
Link this to the conference submission system once it is available.
For ICDM workshops, each submitted paper must be between 5 and 8 pages in length (all figures, tables, and references included), plus 2 additional pages if necessary. All submissions must be original, unpublished work and comply with the double-blind review policy. Accepted papers will be published in the joint ICDM workshop proceedings and archived in IEEE Xplore.
To be announced
