Published in CVPR Workshop on Video Large Language Models, 2026
This workshop paper introduces a diagnostic framework for finding where video-language models make reasoning mistakes and localizing the evidence tied to those failures.
Published in CVPR, 2026
This paper presents a sparse video understanding framework for selecting compact visual evidence while preserving reasoning performance in video-language tasks.
Published in CVPR Workshop on 4D Vision, 2026
This paper evaluates fine-grained physical consistency in 4D scenes generated from video diffusion, connecting visual motion to physics-grounded structure.
Published in ICML, 2026
This paper introduces adversarial co-evolution training for multi-agent reinforcement learning, using safety-driven competition to shape more robust agent behavior.
Published in arXiv, 2026
This paper uses reinforcement learning to refine text-to-video prompts so generated videos better satisfy physical plausibility constraints.
Published in NeurIPS, 2025
This paper frames antibody design as Pareto-optimal energy alignment, balancing binding objectives with nature-like antibody properties.
Published in ICML AI for Math Workshop, 2025
This paper shows how simple feedback can elicit multi-turn reasoning in language models, prompting models to revise and improve solutions over successive attempts.
Published in CPAL, 2025
This paper develops adaptive batch-size schedules for distributed language-model training under data and model parallelism to improve training efficiency.
Published in Information Systems, 2025
This paper combines generative tags with hybrid structural and semantic features to improve schema matching across heterogeneous data sources.
Published in ICML, 2024
This paper introduces a bi-directional cellular tabular learning model built on generalized sparse modern Hopfield memory for supervised learning on tabular data.
Published in arXiv, 2024
This paper proposes adaptive batch-size strategies for distributed local gradient methods to reduce communication cost while maintaining optimization performance.
Published in OpenReview, 2024
This work proposes a distributed hyperparameter search method for time-series prediction models, coordinating search across candidate configurations efficiently.
Published in NeurIPS (ML4Phy Workshop), 2023
This paper introduces a neuralized-PID Soft Actor-Critic (SAC) algorithm for real-time proton beam intensity control in the Mu2e experiment at Fermilab.
Published in NeurIPS, 2023
The sparse modern Hopfield model is introduced as an extension of the dense modern Hopfield model, offering a one-step approximation of the sparse attention mechanism.
Published in IEEE International Conference on Robot and Human Interactive Communication, 2023
This paper studies open-ended multimodal relational reasoning for video question answering and language-based interactions in video scenes.
Published in ICML, 2023
This paper introduces a programmable feature engineering concept for time series modeling through a feature programming framework, emphasizing the efficient generation of predictive features and demonstrating its effectiveness on various time series datasets.
Published in Biomedical Optics Express, 2022
This paper presents RC-UPerNet, a deep learning algorithm for automated identification of cone and rod photoreceptors in high-resolution retinal images.