Scope and Motivation
The field of generative models is undergoing a period of explosive growth and is driving the new area of machine learning systems. This progress is not happening in isolation. Techniques from computer science, information theory, natural language processing, and computer vision have significantly influenced generative models. In turn, these models are giving back by improving established concepts in those very fields. Information theory concepts like compression have found new applications in making generative models more efficient, while generative models themselves are being used to develop novel communication codes. While both machine learning and information theory boast rich histories, the sheer scale and capabilities of these models raise a multitude of new theoretical and algorithmic questions.
Given the rapid recent developments of ideas in generative modeling and its importance in modern day systems, this special issue explores the recent developments in generative models from an information theoretic perspective. The goal of this special issue is not only to capture the recent exciting developments but also attempt to frame important research questions in these domains for the coming years. Therefore we invite expository articles on the following topics of interest (not a comprehensive or exclusive list):
- Application of generative models to compression, codes, and compressed sensing
- Compression of generative models
- Language model alignment
- In-context learning
- Watermarking
- Mechanistic interpretability
- Robustness
- Data efficiency
Full Call for Papers
The full call for papers is available at the BITS website.
BITS Submission Instructions
We will follow the BITS two-stage submission process outlined below and described in BITS Information for Authors at www.itsoc.org/bits/information-authors
Revised Relevant Dates
White paper submission: November 15, 2024
White paper decision: January 17, 2025
Full manuscript submission: March 14, 2025
Full manuscript decision: May 16, 2025
Manuscript final version: June 2025
Special Issue publication: July 2025