---
date: 2026-06-29
phase: 4
headline: More evals + paper figure work
---

## What we did
- Continued eval runs (additional conditions / extended sweep beyond Friday's batch).
- Paper figure work — compiling Fig 4 from the eval results, iterating on the method-diagram figure, and shaping the figure story around the CupSpill 100% result and the data-efficiency framing.

## Next steps
- Resolve the figure-anchor question for the YAM dexterous task (CupSpill vs egg pick-place vs both in Fig).
- Continue eval coverage as needed for OOD-pos / OOD-view evidence on CupSpill.