Science of AI

23 items tagged

Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers

Conference on Computer Vision and Pattern Recognition (CVPR) 2026 (highlight, top 3%)

🌟 Highlight — top 3%

Binxu Wang, Jingxuan Fan, Xu Pan

PDF Diffusion Interpretability Mechanistic Interpretability Computer Vision Science of AI

Where the Score Lives: A Wavelet View of Diffusion

International Conference on Artificial Intelligence and Statistics (AISTATS) 2026

Emma Lucia Byrnes Finn, Binxu Wang, T. Anderson Keller, Demba E. Ba

Diffusion Theory Science of AI

Matching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-training

under review at COLM 2026

William Hoy, Binxu Wang, Xu Pan

PDF LLM Reinforcement Learning Science of AI Theory

Differentiable Faithfulness Alignment for Cross-Model Circuit Transfer

under review at COLM 2026

Shun Shao, Binxu Wang, Shay B Cohen, Yonatan Belinkov

Interpretability Mechanistic Interpretability Science of AI LLM

A Random Matrix Perspective on the Consistency of Diffusion Models

arXiv preprint arXiv:2602.02908 (under review at ICML)

Binxu Wang, Jacob A. Zavatone-Veth, Cengiz Pehlevan

PDF Diffusion Theory Science of AI

When Rule Learning Breaks: Diffusion Fails to Learn Parity of Many Bits

Workshop on Structured Probabilistic Inference & Generative Modeling, NeurIPS 2025 (oral)

🎤 Oral presentation

Binxu Wang, Emma Lucia Byrnes Finn, Bingbin Liu

Diffusion Science of AI Generative Model

An analytical theory of power law spectral bias in the learning dynamics of diffusion models

NeurIPS 2025 (spotlight, top 3%) — arXiv:2503.03206

🌟 Spotlight — top 3%

Binxu Wang, Cengiz Pehlevan

PDF Cite Diffusion Learning Dynamics Theory Science of AI

The Unreasonable Effectiveness of Gaussian Score Approximation for Diffusion Models and its Applications

Transactions on Machine Learning Research (arXiv:2412.09726)

Binxu Wang, John J Vastola

PDF Cite Code 📝 BLOG Diffusion Science of AI Theory

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

Binxu Wang, Jiaqi Shang, Haim Sompolinsky

PDF Cite Diffusion NeuroAI Generative Model Science of AI

Do diffusion models generalize on abstract rules for reasoning?

2024 Conference on Cognitive Computational Neuroscience

Binxu Wang, Jiaqi Shang, Haim Sompolinsky

Cite Diffusion NeuroAI Generative Model Science of AI

Understanding Learning Dynamics of Neural Representations via Feature Visualization at Scale

NeurIPS 2023 UniReps Workshop: the First Workshop on Unifying Representations in Neural Models

Chandana Kuntala, Carlos R Ponce, Deepak Kumar Sharma, Binxu Wang

Cite Learning Dynamics NeuroAI Science of AI Self-supervised learning Computer Vision

The hidden linear structure in score-based models and its application

NeurIPS 2023 Workshop on Diffusion Models (arXiv:2311.10892)

Binxu Wang, John J Vastola

PDF Cite Diffusion Geometry Generative Model Science of AI Theory

Diffusion models generate images like painters: an analytical theory of outline first, details later

arXiv preprint arXiv:2303.02490

Binxu Wang, John J Vastola

PDF Cite Diffusion Geometry Generative Model Science of AI Theory

Science of AI

Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers

Where the Score Lives: A Wavelet View of Diffusion

Matching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-training

Differentiable Faithfulness Alignment for Cross-Model Circuit Transfer

A Random Matrix Perspective on the Consistency of Diffusion Models

When Rule Learning Breaks: Diffusion Fails to Learn Parity of Many Bits

An analytical theory of power law spectral bias in the learning dynamics of diffusion models

The Unreasonable Effectiveness of Gaussian Score Approximation for Diffusion Models and its Applications

Diverse capability and scaling of diffusion and auto-regressive models when learning abstract rules

Do diffusion models generalize on abstract rules for reasoning?

Understanding Learning Dynamics of Neural Representations via Feature Visualization at Scale

The hidden linear structure in score-based models and its application

Diffusion models generate images like painters: an analytical theory of outline first, details later

Computer Vision and Pattern Recognition Conference (CVPR 2026) — Tutorial

Harvard University — CS Systems Seminar

Harvard University — Demba Ba's Class, Introduction to Diffusion Models

University of Missouri-Columbia — Geometry & Topology Seminar

University of Chicago — NeuroML Workshop

Peking University (Shenzhen) — 1st International Young Scholars Forum, School of AI for Science

Renormalization for AI Safety Workshop — Invited Talk

Boston Diffusion Day — Invited Talk

Poster at NEMI 2025 Workshop on Circuit underlying relational object generation in DiT

IAIFI Summer Workshop Talk

Related Tags