Most present vision-language-action (VLA) fashions—similar to OpenVLA, π0, RT-2, and RDT-1B—are “monolithic.” This implies they generate uncooked motor instructions or very quick sequences of actions, with out organizing behaviors into reusable, well-defined abstractions. Consequently, these fashions carry out poorly on long-horizon (multi-step) duties, and it’s tough to interpret what they’ve realized. Current approaches for locating expertise typically keep away from the core drawback of deciding when two motion sequences are “behaviorally equal.” For instance, AtomicVLA and AtomSkill group motion sequences by clustering their contrastive embeddings. In distinction, BLADE and LRLL depend on a big language mannequin (LLM) to evaluate whether or not two sequences are equal, however these LLMs should not calibrated to the robotic’s personal dynamics. We introduce REFACTOR-VLA, a system that learns reusable expertise utilizing a “wake/sleep” structure. Within the sleep section, the system clusters segments of motor packages utilizing a Behavioral-Equivalence Kernel (BEK). This BEK is predicated on the outcomes of rolling out actions in a realized latent world mannequin, Mφ. Within the wake section, the system generates typed lambda phrases (easy, structured packages) from a vocabulary impressed by the Hindley–Milner sort system. These lambda phrases are then utilized by a library-conditioned rectified-flow motion decoder to provide actions. Solely abstractions that go each a Minimal Description Size (MDL) criterion and a return-preservation gate are accepted as expertise. To coach REFACTOR-VLA, we use a three-phase schedule: • Section A (World-model warmup): The latent world mannequin Mφ is skilled. • Section B (Wake-phase coverage optimization): The coverage that makes use of the library of expertise is optimized. • Section C (Sleep-phase ability discovery): The system clusters motion fragments into reusable expertise. We evaluated REFACTOR-VLA on the complete LIBERO benchmark suite. Our outcomes present two major findings. First, merely rising the dimensions of the world mannequin—from 188 million to 430 million parameters—worsened efficiency on 4 out of 4 benchmark suites, disproving the concept that simply making the world mannequin greater at all times helps. Second, altering the coaching goal makes a giant distinction: including an auxiliary supervised contrastive loss (particularly, InfoNCE loss) throughout the world-model warmup (Section A) vastly improved the standard of ability clustering within the sleep section (Section C). We measured this utilizing Normalized Mutual Data (NMI) underneath n = 3 multi-seeding: • Object suite: 0.462 ± 0.021 • Spatial suite: 0.867 ± 0.025 • Purpose suite: 0.915 ± 0.013 • LIBERO-10 suite: 0.754 ± 0.010
