Publications

You can also find my articles on my Google Scholar profile.

Journal Articles

Skip-It? Theoretical Conditions for Layer Skipping in Vision-Language Models

Under Review for ICML 2026. Paper presents a unified framework for analyzing model efficiency improvement methods such as token pruning and layer skipping. We then use this framework and do a case-study on layer-skipping, theoretically grounding its results.

Recommended citation: Hartman, M.*, Jayaraman V.A.*, Choraria, M., Bhimaraju, A., & Varshney, L.R. (2025). Skip-It? Theoretical Conditions for Layer Skipping in Vision-Language Models. Preprint https://www.arxiv.org/abs/2509.25584
Download Paper

Planning Model based on Projection Methodology Bayesian Discrete Extended (PM2-BDE)

Not yet published, undergoing internal review. Improves existing methods for Reliability Growth Planning and system reliability estimation. Manuscript available upon request

Recommended citation: Bernstein, H. C., Bindel, S. R., McKibben, M. A., & Jayaraman, V. A. (2024). Planning Model based on Projection Methodology Bayesian Discrete Extended (PM2-BDE)

SwitchCIT: Switching for Continual Instruction Tuning of Large Language Models

Accepted into Journal of Special Topics in Signal Processing (JSTSP). Paper on utilizing a Switch Network to alleviate catastrophic forgetting in LLMs

Recommended citation: Wu, X., Hartman, M.*, Jayaraman, V. A.*, & Varshney, L. R. (2024, July 16). SwitchCIT: Switching for continual instruction tuning of large language models. arXiv.org. https://arxiv.org/abs/2407.11780
Download Paper

Conference Papers

Unmasking the functionality of the early layers in VLMs

Presented at the eXCV Workshop at ICCV 2025. Interpretability work on determining how vision is transformed in the early layers in VLMs.

Recommended citation: Hartman, M.*, Jayaraman, V.A.*, Choraria, M., Bhimaraju, A., & Varshney, L. R. (2025). Unmasking the functionality of early layers in VLMs
Download Paper