Alexey Zaytsev
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Alexey Zaytsev has a total of 139 co-authors. Here are the top co-authors by number of shared publications:
- Evgeny Burnaev
Skolkovo Institute of Science and Technology
15 shared publications - Evgenia Romanenkova
Skolkovo Institute of Science and Technology (Skoltech), 121205 Moscow, Russia
6 shared publications Abdallah AlShehri
2 shared publicationsAlina Rogulina
2 shared publicationsAnuar Shakirov
2 shared publications- Dmitry Anikin
Applied AI Institute, Moscow, Russian Federation and SB AI Lab, Moscow, Russian Federation
2 shared publications
Latest Publications
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 3883-3894
Never Skip a Batch: Dense Learning of Temporal GNNs via Adaptive Pseudo-Supervision
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 3883-3894
Open AccessNever Skip a Batch: Dense Learning of Temporal GNNs via Adaptive Pseudo-Supervision
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 8954-8965
Benchmarking on Tasks That Matter: Dataset Selection for Preserving Model Rankings
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 8954-8965
Open AccessBenchmarking on Tasks That Matter: Dataset Selection for Preserving Model Rankings
- 30 July 2026
- Published by Springer Nature in Earth Systems and Environment
- Vol. 10 (10) , 14219-14255
From Spectral Indices to Foundation Models: A Review of The AI Revolution in Satellite-based Environmental Hazard Monitoring
- 30 July 2026
- Published by Springer Nature in Earth Systems and Environment
- Vol. 10 (10) , 14219-14255
From Spectral Indices to Foundation Models: A Review of The AI Revolution in Satellite-based Environmental Hazard Monitoring
- Open Access
- 20 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 77279-77301
Looking Around You: External Information Enhances Representations for Event Sequences
Representation learning produces models in different domains, such as store purchases, client transactions, and general people’s behavior. However, such models for event sequences usually process each sequence in isolation, ignoring context from those that co-occur in time. This limitation is particularly problematic in domains with fast-evolving conditions, like finance and e-commerce, or when certain sequences lack recent events. We develop a method that aggregates information from multiple user representations, augmenting a specific user’s representation in a setting with multiple co-occurring event sequences, achieving better quality than processing each sequence independently. Our study considers diverse aggregation approaches, ranging from simple pooling techniques to Learnable attention aggregation, that can highlight more complex information flow among other users. The proposed methods operate on top of an existing encoder and support its efficient fine-tuning. Across nine diverse event sequence datasets (finance, e-commerce, entertainment, etc.) and downstream tasks, Learnable attention improves metric scores, both with and without fine-tuning, while mean pooling yields a smaller but still significant gain.
- 20 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 77279-77301
Open AccessLooking Around You: External Information Enhances Representations for Event Sequences
Representation learning produces models in different domains, such as store purchases, client transactions, and general people’s behavior. However, such models for event sequences usually process each sequence in isolation, ignoring context from those that co-occur in time. This limitation is particularly problematic in domains with fast-evolving conditions, like finance and e-commerce, or when certain sequences lack recent events. We develop a method that aggregates information from multiple user representations, augmenting a specific user’s representation in a setting with multiple co-occurring event sequences, achieving better quality than processing each sequence independently. Our study considers diverse aggregation approaches, ranging from simple pooling techniques to Learnable attention aggregation, that can highlight more complex information flow among other users. The proposed methods operate on top of an existing encoder and support its efficient fine-tuning. Across nine diverse event sequence datasets (finance, e-commerce, entertainment, etc.) and downstream tasks, Learnable attention improves metric scores, both with and without fine-tuning, while mean pooling yields a smaller but still significant gain.
- Open Access
- 15 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 18868-18880
Holistic Uncertainty Estimation for Open-Set Recognition
Accurate uncertainty estimation is a critical challenge in open-set recognition, where a probe biometric sample may belong to an unknown identity. It can be addressed through sample quality estimation via probabilistic embeddings. However, the low variance of probabilistic embedding only partly implies a low identification error probability: an embedding of a sample could be close to several classes in a gallery, thus yielding high uncertainty despite high sample quality. We propose HolUE—a holistic uncertainty estimation method based on a Bayesian probabilistic model; it is aware of two sources of ambiguity in the open-set recognition system: (1) the gallery uncertainty caused by overlapping classes and (2) the uncertainty of embeddings. Challenging open-set recognition datasets, such as IJB-C for the image domain and VoxBlink for the audio domain, serve as a testbed for our method. We also provide a new open-set recognition protocol for the identification of whales and dolphins. In all cases, HolUE better identifies recognition errors than alternative uncertainty estimation methods, including those based solely on sample quality.
- 15 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 18868-18880
Open AccessHolistic Uncertainty Estimation for Open-Set Recognition
Accurate uncertainty estimation is a critical challenge in open-set recognition, where a probe biometric sample may belong to an unknown identity. It can be addressed through sample quality estimation via probabilistic embeddings. However, the low variance of probabilistic embedding only partly implies a low identification error probability: an embedding of a sample could be close to several classes in a gallery, thus yielding high uncertainty despite high sample quality. We propose HolUE—a holistic uncertainty estimation method based on a Bayesian probabilistic model; it is aware of two sources of ambiguity in the open-set recognition system: (1) the gallery uncertainty caused by overlapping classes and (2) the uncertainty of embeddings. Challenging open-set recognition datasets, such as IJB-C for the image domain and VoxBlink for the audio domain, serve as a testbed for our method. We also provide a new open-set recognition protocol for the identification of whales and dolphins. In all cases, HolUE better identifies recognition errors than alternative uncertainty estimation methods, including those based solely on sample quality.
- Open Access
- 8 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 210741-210757
COTIC: Embracing Non-Uniformity in Event Sequence Data via Multilayer Continuous Convolution
Irregular event streams are common in domains such as finance, healthcare, and e-commerce, where the underlying data-generating processes are often highly complex and vary widely in scale. Models must adapt their temporal expressiveness while maintaining training stability. Most neural point-process models interpolate intensity between events using exponential or linear bridges. These parametric assumptions oversimplify temporal dynamics, limit expressiveness, and increase computation complexity. This leads us to the presentation of COTIC. It is a continuous-time convolutional network that models intensity functions directly from raw timestamps. It uses a stack of dilated continuous-convolution layers. The architecture avoids resampling or parametric interpolation. It scales linearly with sequence length. It also captures long-range dependencies through the dilation mechanism. For the first time, we propose a layer-wise propagation and training scheme. This approach enables stable optimization of deep continuous-time convolutional architectures for event sequences. Across seven open benchmarks, COTIC achieves an average rank of 1.5. It outperforms Transformer-based and RNN-based baselines. It is the first fully continuous-convolutional framework to combine long-term memory with stable multilayer training. Moreover, its encoder produces high-quality transferable embeddings. COTIC provides a general, scalable foundation for modeling irregular event streams across domains. Code and configuration files are available at https://github.com/VladislavZh/COTIC.
- 8 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 210741-210757
Open AccessCOTIC: Embracing Non-Uniformity in Event Sequence Data via Multilayer Continuous Convolution
Irregular event streams are common in domains such as finance, healthcare, and e-commerce, where the underlying data-generating processes are often highly complex and vary widely in scale. Models must adapt their temporal expressiveness while maintaining training stability. Most neural point-process models interpolate intensity between events using exponential or linear bridges. These parametric assumptions oversimplify temporal dynamics, limit expressiveness, and increase computation complexity. This leads us to the presentation of COTIC. It is a continuous-time convolutional network that models intensity functions directly from raw timestamps. It uses a stack of dilated continuous-convolution layers. The architecture avoids resampling or parametric interpolation. It scales linearly with sequence length. It also captures long-range dependencies through the dilation mechanism. For the first time, we propose a layer-wise propagation and training scheme. This approach enables stable optimization of deep continuous-time convolutional architectures for event sequences. Across seven open benchmarks, COTIC achieves an average rank of 1.5. It outperforms Transformer-based and RNN-based baselines. It is the first fully continuous-convolutional framework to combine long-term memory with stable multilayer training. Moreover, its encoder produces high-quality transferable embeddings. COTIC provides a general, scalable foundation for modeling irregular event streams across domains. Code and configuration files are available at https://github.com/VladislavZh/COTIC.
- Open Access
- 20 November 2025
- Published by Elsevier in Ecological Informatics
- Vol. 92
Foundation for unbiased cross-validation of spatio-temporal models for Species Distribution Modeling
- 20 November 2025
- Published by Elsevier in Ecological Informatics
- Vol. 92
Open AccessFoundation for unbiased cross-validation of spatio-temporal models for Species Distribution Modeling
- Open Access
- 3 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 193893-193908
Concealed Adversarial Attacks on Neural Networks for Sequential Data
The emergence of deep learning has enabled the widespread application of neural networks in time series analysis across domains such as finance and medicine. Despite their effectiveness, these models remain vulnerable to adversarial attacks: small targeted perturbations in input data can significantly alter a classifier’s outputs. However, even formally weak attacks in the time series domain become easily detected by the human eye or a simple detector model. In this work, a concealed adversarial attack is introduced for various time series models. The approach generates realistic perturbations that are difficult for both humans and model-based discriminators to detect by jointly maximizing classifier loss and discriminator loss. A specialized training strategy is employed to enhance the discriminator’s ability to generalize across different attack strengths. Evaluation on eight benchmark datasets from the University of California Riverside (UCR) and University of East Anglia (UEA) archives, using recurrent, convolutional, state-space, and transformer-based architectures, demonstrates that the proposed method achieves a favorable trade-off between concealability and attack effectiveness. The findings highlight the growing challenge of designing robust time series models, emphasizing the need for improved defenses against stealthy and effective attacks.
- 3 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 193893-193908
Open AccessConcealed Adversarial Attacks on Neural Networks for Sequential Data
The emergence of deep learning has enabled the widespread application of neural networks in time series analysis across domains such as finance and medicine. Despite their effectiveness, these models remain vulnerable to adversarial attacks: small targeted perturbations in input data can significantly alter a classifier’s outputs. However, even formally weak attacks in the time series domain become easily detected by the human eye or a simple detector model. In this work, a concealed adversarial attack is introduced for various time series models. The approach generates realistic perturbations that are difficult for both humans and model-based discriminators to detect by jointly maximizing classifier loss and discriminator loss. A specialized training strategy is employed to enhance the discriminator’s ability to generalize across different attack strengths. Evaluation on eight benchmark datasets from the University of California Riverside (UCR) and University of East Anglia (UEA) archives, using recurrent, convolutional, state-space, and transformer-based architectures, demonstrates that the proposed method achieves a favorable trade-off between concealability and attack effectiveness. The findings highlight the growing challenge of designing robust time series models, emphasizing the need for improved defenses against stealthy and effective attacks.
- Open Access
Selective Adversarial Attacks on LLM Benchmarks
Benchmarking outcomes increasingly govern trust, selection, and deployment of LLMs, yet these evaluations remain vulnerable to semantically equivalent adversarial perturbations. Prior work on adversarial robustness in NLP has emphasized text attacks that affect many models equally, leaving open the question of whether it is possible to selectively degrade or enhance performance while minimally affecting other models. We formalize this problem and study selective adversarial attacks on MMLU - a widely used benchmark designed to measure a language model's broad general knowledge and reasoning ability across different subjects. Using canonical attacks integrated into TextAttack framework, we introduce a protocol for selectivity assessment, develop a custom constraint to increase selectivity of attacks and propose a surrogate-LLM pipeline that generates selective perturbations. Empirically, we find that selective adversarial attacks exist and can materially alter relative rankings, challenging the fairness, reproducibility, and transparency of leaderboard-driven evaluation. Our results motivate perturbation-aware reporting and robustness diagnostics for LLM evaluation and demonstrate that even subtle edits can shift comparative judgments.
Open AccessSelective Adversarial Attacks on LLM Benchmarks
Benchmarking outcomes increasingly govern trust, selection, and deployment of LLMs, yet these evaluations remain vulnerable to semantically equivalent adversarial perturbations. Prior work on adversarial robustness in NLP has emphasized text attacks that affect many models equally, leaving open the question of whether it is possible to selectively degrade or enhance performance while minimally affecting other models. We formalize this problem and study selective adversarial attacks on MMLU - a widely used benchmark designed to measure a language model's broad general knowledge and reasoning ability across different subjects. Using canonical attacks integrated into TextAttack framework, we introduce a protocol for selectivity assessment, develop a custom constraint to increase selectivity of attacks and propose a surrogate-LLM pipeline that generates selective perturbations. Empirically, we find that selective adversarial attacks exist and can materially alter relative rankings, challenging the fairness, reproducibility, and transparency of leaderboard-driven evaluation. Our results motivate perturbation-aware reporting and robustness diagnostics for LLM evaluation and demonstrate that even subtle edits can shift comparative judgments.