Yujin Song

3 papers · 2024–2025 · 3 venues · across top CS/AI venues

Authors are matched by name, so researchers who share a name ("Yang Liu", "Wei Wang") appear as a single profile and their paper counts, venues and achievements are combined. How identity is handled.

Achievements

Jump to papers ↓

🌍 Conference Polyglot · Bronze 🗺️ Taxonomy Completionist (11)

Venues

COLT (1) ICML (1) NeurIPS (1)

Top co-authors

Kazusato Oko (3) Denny Wu (3) Taiji Suzuki (3) Naoki Nishikawa (1)

Research topics

Deep Learning (3) Machine Learning (3) Architectures (3) Core Methods (3) Optimization & Theory · Machine Learning (3) Artificial Intelligence (1) Core AI (1) Learning Types · Machine Learning (1) Learning Types · Deep Learning (1) Optimization & Theory · Deep Learning (1)

Keywords

single-index model (3) in-context learning (2) gradient descent (2) representation learning (2) nonlinear regression (1) additive model (1) low-dimensional function (1) information exponent (1) two-layer neural network (1) pre-trained transformer (1) feature vector (1) transformer model (1) statistical query lower bound (1) nonlinear function class (1) single-index function (1) ridge function (1) gradient descent training (1) transformer architecture (1) softmax self-attention (1) sample complexity (1)

Papers

Nonlinear transformers can perform inference-time feature learning ICML 2025 Learning sum of diverse features: computational hardness and efficient gradient-based training for ridge combinations COLT 2024 Pretrained Transformer Efficiently Learns Low-Dimensional Target Functions In-Context NeurIPS 2024