Scholar
David Baek
Google Scholar ID: ArkB1_MAAAAJ
PhD Student, MIT
Citations & Impact
Citations
51
H-index
4
i10-index
1
Publications
7
Co-authors
5
list available
Publications
Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
- “Harmonic Loss Trains Interpretable AI Models” (arXiv, 2025): introduced harmonic loss as an alternative to cross-entropy loss; Twitter post received 1.1M views
- “The Geometry of Concepts: Sparse Autoencoder Feature Structure” (arXiv, 2024): analyzed geometric structures of SAE features at multiple scales; Twitter post received 300K views
- “Generalization from Starvation: Hints of Universality in LLM Knowledge Graph Learning” (arXiv, 2024): explored universality in LLM knowledge representations
- “GenEFT: Understanding Statics and Dynamics of Model Generalization via Effective Theory” (ICLR Workshop, 2024): proposed an effective field theory framework for neural network generalization
- “Design of scalable superconducting quantum circuits using flip-chip assembly” (IEEE Transactions on Applied Conductivity, 2022)
- Class project: “Filtering Data to Improve NeRF for Driving Scenes” (2024): proposed a filter-based preprocessing strategy for NeRF in autonomous driving
- Korean Military Open Source Hackathon (2020): developed an automated leave management system for Korean soldiers
- Mobile Computing course project (2019): built “Wanderlust,” a global hiking community mobile app
