The rapid development of large language models (LLMs) calls for solid theoretical foundations to tackle core challenges including scalability, interpretability, and robustness. Information theory provides powerful analytical frameworks for AI data processing and system optimization, yet its integration into modern LLM research remains insufficiently explored.

This Special Issue aims to fill this gap, welcoming innovative theoretical research and novel algorithms that leverage information theory to advance the capability and interpretability of LLMs, promoting interdisciplinary innovation in this fast-growing field.

* 15 November 2026: Manuscript submission deadline * Rolling publication: Accepted papers will be published online continuously upon approval

We welcome original research and review papers covering (but not limited to):

1. Information-theoretic foundations of LLMs 2. Memory and retrieval mechanisms in LLMs 3. Reinforcement learning and LLM reasoning 4. Generalization bounds for LLM training 5. Uncertainty quantification and hallucination detection in LLMs 6. Multimodal LLM information fusion 7. Energy-efficient LLMs via information-theoretic optimization 8. LLM watermarking technologies 9. Large language model compression algorithms

On behalf of the Guest Editors: Dr. Xueyan Niu, Prof. Dr. Jun Chen, Dr. Bo Bai