Projects
- Agents Tried to Cheat at GeoGuessr ★★★★☆
- Extending "Decomposing and Measuring Evaluation Awareness" (Li et al. 2026) ★★★★☆
- Models are Awful at Controlling Their CoT ★★★☆☆
- Natural Deduction Takehome ★★★☆☆
- Pretraining ablations, week of September 21, 2026 ★★☆☆☆
- The combined model, and where its gains come from ★☆☆☆☆