Science of AI
21 items tagged
(2026).
Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers.
Conference on Computer Vision and Pattern Recognition (CVPR) 2026 (highlight, top 3%).
(2026).
Where the Score Lives: A Wavelet View of Diffusion.
International Conference on Artificial Intelligence and Statistics (AISTATS) 2026.
(2026).
Matching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-training.
under review at COLM 2026.
(2026).
Differentiable Faithfulness Alignment for Cross-Model Circuit Transfer.
under review at COLM 2026.
(2025).
When Rule Learning Breaks: Diffusion Fails to Learn Parity of Many Bits.
Workshop on Structured Probabilistic Inference & Generative Modeling, NeurIPS 2025 (oral).
(2025).
An analytical theory of power law spectral bias in the learning dynamics of diffusion models.
NeurIPS 2025 (spotlight, top 3%) — arXiv:2503.03206.
(2025).
A Random Matrix Perspective on the Consistency of Diffusion Models.
arXiv preprint arXiv:2602.02908 (under review at ICML).
(2024).
The Unreasonable Effectiveness of Gaussian Score Approximation for Diffusion Models and its Applications.
Transactions on Machine Learning Research (arXiv:2412.09726).
(2024).
Diverse capability and scaling of diffusion and auto-regressive models when learning abstract rules.
The First Workshop on System-2 Reasoning at Scale, NeurIPS'24.
(2024).
Do diffusion models generalize on abstract rules for reasoning?.
2024 Conference on Cognitive Computational Neuroscience.
(2023).
Understanding Learning Dynamics of Neural Representations via Feature Visualization at Scale.
NeurIPS 2023 UniReps Workshop: the First Workshop on Unifying Representations in Neural Models.
(2023).
The hidden linear structure in score-based models and its application.
NeurIPS 2023 Workshop on Diffusion Models (arXiv:2311.10892).
(2023).
Diffusion models generate images like painters: an analytical theory of outline first, details later.
arXiv preprint arXiv:2303.02490.