We are a research lab dedicated to advancing the theoretical and empirical understanding of deep learning. Our work spans from analyzing learning dynamics and representation formation to investigating the fundamental reasons behind generalization in neural networks.
DLFundamentals
Popular repositories Loading
-
directional-nc
directional-nc PublicOfficial code for "Directional Neural Collapse Explains Few-Shot Transfer in Self-Supervised Learning", ICML 2026
Jupyter Notebook 12
-
ssl-vs-sl-pt2
ssl-vs-sl-pt2 PublicOfficial code for "On the Alignment Between Supervised and Self-Supervised Contrastive Learning", ICLR 2026
Python 8
-
distribution-aware-algorithm-design
distribution-aware-algorithm-design PublicOfficial code for "Distribution-Aware Algorithm Design with LLM Agents"
-
ssl-vs-sl-pt1
ssl-vs-sl-pt1 PublicOfficial code for "Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning", NeurIPS 2025
Python 7
-
Repositories
Showing 10 of 13 repositories
- dlfundamentals.github.io Public
- distribution-aware-algorithm-design Public
Official code for "Distribution-Aware Algorithm Design with LLM Agents"
- swebench-orchestration Public
Code for "Agentic Systems as Boosting Weak Reasoning Models" — verifier-backed committee search that boosts weak reasoning models to frontier-level SWE-bench performance at inference time!
- blog Public
- directional-nc Public
Official code for "Directional Neural Collapse Explains Few-Shot Transfer in Self-Supervised Learning", ICML 2026
- RepX Public
- directional-neural-collapse Public
- ssl-vs-sl-pt2 Public
Official code for "On the Alignment Between Supervised and Self-Supervised Contrastive Learning", ICLR 2026
Top languages
Loading…
Most used topics
Loading…