About

Anna Choromanska is an Associate Professor in the Department of Electrical and Computer Engineering at NYU Tandon School of Engineering. She is affiliated with several NYU centers, including the NYU Center for Data Science, the NYU Center for Urban Science and Progress, the Center for Advanced Technology in Communications, and the Connected Cities with Smart Transportation Center. Her main research focus is deep learning, particularly understanding how machines acquire knowledge, optimizing and training deep learning models, and analyzing large datasets. Her work aims to develop fundamental insights into deep learning optimization and generalization, which are critical for designing efficient, accurate, and scalable AI systems. Her research involves multidisciplinary approaches, combining experimental and theoretical methods, with applications in autonomous driving and large dataset analysis. Prof. Choromanska has contributed to industry through her work being used in production by companies such as Facebook, Baidu, and NVIDIA, and she has been active in the academic community as a speaker, editor, organizer, and reviewer. She holds a Ph.D. and M.Phil. from Columbia University and a master's degree from Warsaw University of Technology. Her prior experience includes postdoctoral work at NYU's Courant Institute, internships at Microsoft Research, IBM T.J. Watson Research Center, and various visiting scholar positions. She has received numerous awards, including the NSF CAREER Award, the Alfred P. Sloan Fellowship, and IBM Global University Program Academic Awards. Beyond her scientific pursuits, she is also a talented pianist, a dancer, and an enthusiast of painting and fashion design.

Research topics

• Artificial Intelligence
• Computer Science
• Computer Security
• Machine Learning
• Mathematics
• Real-time computing
• Engineering
• Control engineering

Selected publications

• Zero-Shot Cross-City Generalization in End-to-End Autonomous Driving: Self-Supervised versus Supervised Representations

  arXiv (Cornell University) · 2026-03-12

  preprintOpen accessSenior author

  End-to-end autonomous driving models are typically trained on multi-city datasets using supervised ImageNet-pretrained backbones, yet their ability to generalize to unseen cities remains largely unexamined. When training and evaluation data are geographically mixed, models may implicitly rely on city-specific cues, masking failure modes that would occur under real domain shifts when generalizing to new locations. In this work we investigate zero-shot cross-city generalization in end-to-end trajectory planning and ask whether self-supervised visual representations improve transfer across cities. We conduct a comprehensive study by integrating self-supervised backbones (I-JEPA, DINOv2, and MAE) into planning frameworks. We evaluate performance under strict geographic splits on nuScenes in the open-loop setting and on NAVSIM in the closed-loop evaluation protocol. Our experiments reveal a substantial generalization gap when transferring models relying on traditional supervised backbones across cities with different road topologies and driving conventions, particularly when transferring from right-side to left-side driving environments. Self-supervised representation learning reduces this gap. In open-loop evaluation, a supervised backbone exhibits severe inflation when transferring from Boston to Singapore (L2 displacement ratio 9.77x, collision ratio 19.43x), whereas domain-specific self-supervised pretraining reduces this to 1.20x and 0.75x respectively. In closed-loop evaluation, self-supervised pretraining improves PDMS by up to 4 percent for all single-city training cities. These results show that representation learning strongly influences the robustness of cross-city planning and establish zero-shot geographic transfer as a necessary test for evaluating end-to-end autonomous driving systems.

  PublisherDOI

• Self-Supervised JEPA-based World Models for LiDAR Occupancy Completion and Forecasting

  arXiv (Cornell University) · 2026-02-13

  articleOpen accessSenior author

  Autonomous driving, as an agent operating in the physical world, requires the fundamental capability to build \textit{world models} that capture how the environment evolves spatiotemporally in order to support long-term planning. At the same time, scalability demands learning such models in a self-supervised manner; \textit{joint-embedding predictive architecture (JEPA)} enables learning world models via leveraging large volumes of unlabeled data without relying on expensive human annotations. In this paper, we propose \textbf{AD-LiST-JEPA}, a self-supervised world model for autonomous driving that predicts future spatiotemporal evolution from LiDAR data using a JEPA framework. We evaluate the quality of the learned representations through a downstream LiDAR-based occupancy completion and forecasting (OCF) task, which jointly assesses perception and prediction. Proof of concept experiments show better OCF performance with pretrained encoder after JEPA-based world model learning.

  PublisherOA PDF

• Self-Supervised JEPA-based World Models for LiDAR Occupancy Completion and Forecasting

  Open MIND · 2026-02-13

  preprintSenior author

  Autonomous driving, as an agent operating in the physical world, requires the fundamental capability to build \textit{world models} that capture how the environment evolves spatiotemporally in order to support long-term planning. At the same time, scalability demands learning such models in a self-supervised manner; \textit{joint-embedding predictive architecture (JEPA)} enables learning world models via leveraging large volumes of unlabeled data without relying on expensive human annotations. In this paper, we propose \textbf{AD-LiST-JEPA}, a self-supervised world model for autonomous driving that predicts future spatiotemporal evolution from LiDAR data using a JEPA framework. We evaluate the quality of the learned representations through a downstream LiDAR-based occupancy completion and forecasting (OCF) task, which jointly assesses perception and prediction. Proof of concept experiments show better OCF performance with pretrained encoder after JEPA-based world model learning.

  DOI

• Zero-Shot Cross-City Generalization in End-to-End Autonomous Driving: Self-Supervised versus Supervised Representations

  ArXiv.org · 2026-03-12

  articleOpen accessSenior author

  End-to-end autonomous driving models are typically trained on multi-city datasets using supervised ImageNet-pretrained backbones, yet their ability to generalize to unseen cities remains largely unexamined. When training and evaluation data are geographically mixed, models may implicitly rely on city-specific cues, masking failure modes that would occur under real domain shifts when generalizing to new locations. In this work we investigate zero-shot cross-city generalization in end-to-end trajectory planning and ask whether self-supervised visual representations improve transfer across cities. We conduct a comprehensive study by integrating self-supervised backbones (I-JEPA, DINOv2, and MAE) into planning frameworks. We evaluate performance under strict geographic splits on nuScenes in the open-loop setting and on NAVSIM in the closed-loop evaluation protocol. Our experiments reveal a substantial generalization gap when transferring models relying on traditional supervised backbones across cities with different road topologies and driving conventions, particularly when transferring from right-side to left-side driving environments. Self-supervised representation learning reduces this gap. In open-loop evaluation, a supervised backbone exhibits severe inflation when transferring from Boston to Singapore (L2 displacement ratio 9.77x, collision ratio 19.43x), whereas domain-specific self-supervised pretraining reduces this to 1.20x and 0.75x respectively. In closed-loop evaluation, self-supervised pretraining improves PDMS by up to 4 percent for all single-city training cities. These results show that representation learning strongly influences the robustness of cross-city planning and establish zero-shot geographic transfer as a necessary test for evaluating end-to-end autonomous driving systems.

  PublisherOA PDF

• Understanding Quantization of Optimizer States in LLM Pre-training: Dynamics of State Staleness and Effectiveness of State Resets

  arXiv (Cornell University) · 2026-03-17

  preprintOpen accessSenior author

  Quantizing optimizer states is becoming an important ingredient of memory-efficient large-scale pre-training, but the resulting optimizer dynamics remain only partially understood. We study low-precision exponential moving average (EMA) optimizer states and show how quantization can cause many nominal updates to round back to the same stored value, making the state effectively stale and slowing adaptation beyond what the nominal decay would suggest. We then develop a simple predictive model of stalling that estimates one-step stalling probabilities and characterizes how stalling builds up over time after the initialization. This perspective provides a mechanistic explanation for why optimizer-state resets help in low precision: once a quantized EMA becomes effectively stale, resetting it can temporarily restore responsiveness. Motivated by this picture, we derive a simple theory-guided method for choosing useful reset periods, showing that in low precision the key question is not only whether resets help, but when they should be applied. Experiments in controlled simulations and LLM pre-training show that suitable reset schedules recover the performance lost to low-precision state storage while substantially reducing optimizer-state memory.

  PublisherDOI

• Understanding Quantization of Optimizer States in LLM Pre-training: Dynamics of State Staleness and Effectiveness of State Resets

  ArXiv.org · 2026-03-17

  articleOpen accessSenior author

  Quantizing optimizer states is becoming an important ingredient of memory-efficient large-scale pre-training, but the resulting optimizer dynamics remain only partially understood. We study low-precision exponential moving average (EMA) optimizer states and show how quantization can cause many nominal updates to round back to the same stored value, making the state effectively stale and slowing adaptation beyond what the nominal decay would suggest. We then develop a simple predictive model of stalling that estimates one-step stalling probabilities and characterizes how stalling builds up over time after the initialization. This perspective provides a mechanistic explanation for why optimizer-state resets help in low precision: once a quantized EMA becomes effectively stale, resetting it can temporarily restore responsiveness. Motivated by this picture, we derive a simple theory-guided method for choosing useful reset periods, showing that in low precision the key question is not only whether resets help, but when they should be applied. Experiments in controlled simulations and LLM pre-training show that suitable reset schedules recover the performance lost to low-precision state storage while substantially reducing optimizer-state memory.

  PublisherOA PDF

• Self-Supervised Representation Learning with Joint Embedding Predictive Architecture for Automotive LiDAR Object Detection

  Proceedings of the AAAI Conference on Artificial Intelligence · 2026-03-14

  articleOpen accessSenior author

  Recently, self‑supervised representation learning relying on vast amounts of unlabeled data has been explored as a pre‑training method for autonomous driving. However, directly applying popular contrastive or generative methods to this problem is insufficient and may even lead to negative transfer. In this paper, we present AD‑L‑JEPA, a novel self‑supervised pre‑training framework with a joint embedding predictive architecture (JEPA) for automotive LiDAR object detection. Unlike existing methods, AD‑L‑JEPA is neither generative nor contrastive. Instead of explicitly generating masked regions, our method predicts Bird’s‑Eye‑View embeddings to capture the diverse nature of driving scenes. Furthermore, our approach eliminates the need to manually form contrastive pairs by employing explicit variance regularization to avoid representation collapse. Experimental results demonstrate consistent improvements on the LiDAR 3D object detection downstream task across the KITTI3D, Waymo, and ONCE datasets, while reducing GPU hours by 1.9×–2.7× and GPU memory by 2.8×–4× compared with the state-of-the-art method Occupancy-MAE. Notably, on the largest ONCE dataset, pre‑training on 100K frames yields a 1.61 mAP gain, better than in case of all the other methods pre‑trained on either 100K or 500K frames, and pre‑training on 500K frames yields a 2.98 mAP gain, better than in case of all the other methods pre‑trained on either 500K or 1M frames. AD‑L‑JEPA constitutes the first JEPA‑based pre‑training method for autonomous driving. It offers better quality, faster, and more GPU‑memory‑efficient self‑supervised representation learning.

  PublisherOA PDFDOI

• Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization

  ArXiv.org · 2025-10-06

  preprintOpen accessSenior author

  The vast majority of modern deep learning models are trained with momentum-based first-order optimizers. The momentum term governs the optimizer's memory by determining how much each past gradient contributes to the current convergence direction. Fundamental momentum methods, such as Nesterov Accelerated Gradient and the Heavy Ball method, as well as more recent optimizers such as AdamW and Lion, all rely on the momentum coefficient that is customarily set to $β= 0.9$ and kept constant during model training, a strategy widely used by practitioners, yet suboptimal. In this paper, we introduce an \textit{adaptive memory} mechanism that replaces constant momentum with a dynamic momentum coefficient that is adjusted online during optimization. We derive our method by approximating the objective function using two planes: one derived from the gradient at the current iterate and the other obtained from the accumulated memory of the past gradients. To the best of our knowledge, such a proximal framework was never used for momentum-based optimization. Our proposed approach is novel, extremely simple to use, and does not rely on extra assumptions or hyperparameter tuning. We implement adaptive memory variants of both SGD and AdamW across a wide range of learning tasks, from simple convex problems to large-scale deep learning scenarios, demonstrating that our approach can outperform standard SGD and Adam with hand-tuned momentum coefficients. Finally, our work opens doors for new ways of inducing adaptivity in optimization.

  PublisherOA PDFDOI

• A Survey of Optimization Methods for Training DL Models: Theoretical Perspective on Convergence and Generalization

  ArXiv.org · 2025-01-24

  preprintOpen accessSenior author

  As data sets grow in size and complexity, it is becoming more difficult to pull useful features from them using hand-crafted feature extractors. For this reason, deep learning (DL) frameworks are now widely popular. The Holy Grail of DL and one of the most mysterious challenges in all of modern ML is to develop a fundamental understanding of DL optimization and generalization. While numerous optimization techniques have been introduced in the literature to navigate the exploration of the highly non-convex DL optimization landscape, many survey papers reviewing them primarily focus on summarizing these methodologies, often overlooking the critical theoretical analyses of these methods. In this paper, we provide an extensive summary of the theoretical foundations of optimization methods in DL, including presenting various methodologies, their convergence analyses, and generalization abilities. This paper not only includes theoretical analysis of popular generic gradient-based first-order and second-order methods, but it also covers the analysis of the optimization techniques adapting to the properties of the DL loss landscape and explicitly encouraging the discovery of well-generalizing optimal points. Additionally, we extend our discussion to distributed optimization methods that facilitate parallel computations, including both centralized and decentralized approaches. We provide both convex and non-convex analysis for the optimization algorithms considered in this survey paper. Finally, this paper aims to serve as a comprehensive theoretical handbook on optimization methods for DL, offering insights and understanding to both novice and seasoned researchers in the field.

  PublisherOA PDFDOI

• OncoReason: Structuring Clinical Reasoning in LLMs for Robust and Interpretable Survival Prediction

  ArXiv.org · 2025-10-20

  preprintOpen accessSenior author

  Predicting cancer treatment outcomes requires models that are both accurate and interpretable, particularly in the presence of heterogeneous clinical data. While large language models (LLMs) have shown strong performance in biomedical NLP, they often lack structured reasoning capabilities critical for high-stakes decision support. We present a unified, multi-task learning framework that aligns autoregressive LLMs with clinical reasoning for outcome prediction on the MSK-CHORD dataset. Our models are trained to jointly perform binary survival classification, continuous survival time regression, and natural language rationale generation. We evaluate three alignment strategies: (1) standard supervised fine-tuning (SFT), (2) SFT with Chain-of-Thought (CoT) prompting to elicit step-by-step reasoning, and (3) Group Relative Policy Optimization (GRPO), a reinforcement learning method that aligns model outputs to expert-derived reasoning trajectories. Experiments with LLaMa3-8B and Med42-8B backbones demonstrate that CoT prompting improves F1 by +6.0 and reduces MAE by 12%, while GRPO achieves state-of-the-art interpretability and predictive performance across BLEU, ROUGE, and BERTScore. We further show that existing biomedical LLMs often fail to produce valid reasoning traces due to architectural constraints. Our findings underscore the importance of reasoning-aware alignment in multi-task clinical modeling and set a new benchmark for interpretable, trustworthy LLMs in precision oncology.

  PublisherOA PDFDOI

Frequent coauthors

• Krzysztof Choromański

  58 shared

• Mariusz Bojarski

  New York University

  50 shared

• Anne Morvan

  39 shared

• Cédric Gouy‐Pailler

  Commissariat à l'Énergie Atomique et aux Énergies Alternatives

  38 shared

• Francois Fagan

  37 shared

• Nourhan Sakr

  American University in Cairo

  37 shared

• Tamás Sarlós

  Google (United States)

  37 shared

• Jamal Atif

  36 shared

Awards & honors

• NSF CAREER Award (2021)
• Alfred P. Sloan Research Fellowship in Computer Science (202…
• IBM Global University Program Academic Award (2021)
• IBM Faculty Award (2020)
• Student Best Paper Award, First Place, 7th Annual Machine Le…