# Apple — REFACTOR-VLA: Unsupervised Library Learning of Typed Motor Programs

- Company: Apple (apple.com)
- Announced: 2026-09-02
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/refactor-vla-motor-programs
- Record: https://forck.live/items/10302-refactor-vla-unsupervised-library-learning-of-typed-motor-programs
- Subject: Machine Learning Research

Apple researchers introduce REFACTOR-VLA, a system that learns reusable motor skills using a wake/sleep architecture with a Behavioral-Equivalence Kernel and typed lambda terms. Evaluated on the LIBERO benchmark, the study finds that increasing world model size from 188M to 430M parameters worsened performance on all four suites, while adding an auxiliary InfoNCE contrastive loss during world-model warmup improved skill clustering quality.

## Evidence

Verbatim from https://machinelearning.apple.com/research/refactor-vla-motor-programs:

> We introduce REFACTOR-VLA, a system that learns reusable skills using a "wake/sleep" architecture. In the sleep phase, the system clusters segments of motor programs using a Behavioral-Equivalence Kernel (BEK).

---

Record: https://forck.live/items/10302-refactor-vla-unsupervised-library-learning-of-typed-motor-programs
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md