Research / published work
Questions, experiments, and evidence from the lab.
Research notes are written like small papers. They keep the question, setup, evidence, limits, and next experiments in the same place.
Research notes
3 notesOctober 2026
Research note
October 2026Endogenous subspace orthogonalization: targeted monosemanticity via LoRA
Ortho-LoRA uses a task-conditioned geometry loss to reduce overlap between adapter channels, with the aim of making model behavior easier to inspect and steer.
Read researchResearch note
September 2026LoRA composition: capability retention without functional composition
A controlled study of when merged adapters retain skills, interfere with one another, and fail to execute an unseen ordered task.
Read researchResearch note
DriftLab: measuring agent adaptation in non-stationary environments
A testbed for studying how agents notice, recover from, and retain changes in the worlds they work in.
Read research