Shikhar Srivastava
I'm a Computer Science PhD student at the University of Rochester, advised by Christopher Kanan.
I work to enable large models to learn continually. My current interests are in memory, learning dynamics of large language models, and continual consolidation in large language models.
I also work with Catherine Arnett at EleutherAI, Dhireesha Kudithipudi at UTSA.
Personal
I'm fascinated by memory. By the phenomena arising from our brains developing collectively through the phylogenetic tree. By old stories and old ideas.
I believe strongly in working towards the benefit of all.
Publications
-
LayerRoPE: Dynamic Depth-wise Magnitude & Angular Superposition
Under review · Actionable Interpretability, COLM 2026 -
No Single Tokenizer Feature Reliably Predicts Downstream Language Model Performance
Under review · EMNLP 2026 -
SHARP: Sleep-based Hierarchical Accelerated Replay for Long Range Non-Stationary Temporal Pattern Recognition[pdf]
CoLLAs 2026 -
Journal of Data-centric Machine Learning Research (DMLR 2026) -
CoLLAs 2025 · Scalable Continual Learning for Lifelong Foundation Models, NeurIPS 2024 -
Benchmarking & Expanding AI Multimodal Approaches, CVPR 2025 · Navigating and Addressing Data Problems for Foundation Models, ICLR 2025 -
Best Paper Award · FAIR, MICCAI 2021
***