Pinned Loading

1. nano-opd nano-opd Public

   Minimal from-scratch on-policy distillation on a single A100: f-divergence advantage design, reproducing the OPD+ correction on GSM8K

   Jupyter Notebook 1

2. Multi-task-Learning-Biomass-Estimation Multi-task-Learning-Biomass-Estimation Public

   Master Thesis - Project, 2025/2026

   Jupyter Notebook 1

3. scaling-law-lab scaling-law-lab Public

   From-scratch Chinchilla scaling laws on a single A100: an IsoFLOP grid over 0.5M-45M-param models on TinyStories predicts a held-out 3x-larger run within 1.29%, and traces the Kaplan-Chinchilla exp…

   Jupyter Notebook 1

4. Physics-Aware-Multi-Task-Learning-for-Forest-and-Biomass-Estimation Physics-Aware-Multi-Task-Learning-for-Forest-and-Biomass-Estimation Public

   Master Thesis - Project, 2nd Part, 2025/2026

   Jupyter Notebook

5. keel-agent-lab keel-agent-lab Public

   Agent runtime with an ablation pipeline that quantifies harness-layer changes

   Python 1