Habilis-$β$: A Fast-Motion and Long-Lasting On-Device Vision-Language-Action Model

2026-02-21

AI Summary

This work addresses the limitations of existing vision-language-action models, which are typically evaluated solely on single-task success rates and thus fail to capture the throughput and long-term reliability required for real-world deployment. To bridge this gap, the authors propose a vision-language-action model tailored for edge-based real-world scenarios, introducing the Productivity-Reliability Plane (PRP) evaluation framework grounded in continuous-operation protocols. Key innovations include language-agnostic pretraining on large-scale play data, cyclic task fine-tuning, phase-adaptive motion planning (ESPADA), rectified flow distillation, and classifier-free guidance. The model achieves 572.6 tasks per hour (TPH) with a mean time between interventions (MTBI) of 39.2 seconds in simulation, and 124 TPH with 137.4 seconds MTBI on real-world logistics tasks—significantly outperforming baselines and establishing state-of-the-art performance on the RoboTwin 2.0 benchmark.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Language and Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

We introduce Habilis-$β$, a fast-motion and long-lasting on-device vision-language-action (VLA) model designed for real-world deployment. Current VLA evaluation remains largely confined to single-trial success rates under curated resets, which fails to capture the fast-motion and long-lasting capabilities essential for practical operation. To address this, we introduce the Productivity-Reliability Plane (PRP), which evaluates performance through Tasks per Hour (TPH) and Mean Time Between Intervention (MTBI) under a continuous-run protocol that demands both high-speed execution and sustained robustness. Habilis-$β$ achieves high performance by integrating language-free pre-training on large-scale play data for robust interaction priors with post-training on cyclic task demonstrations that capture state drift across consecutive task iterations. The system further employs ESPADA for phase-adaptive motion shaping to accelerate free-space transit, utilizes rectified-flow distillation to enable high-frequency control on edge devices, and incorporates classifier-free guidance (CFG) as a deployment-time knob to dynamically balance instruction adherence and learned interaction priors. In 1-hour continuous-run evaluations, Habilis-$β$ achieves strong performance under the PRP metrics, compared to $π_{0.5}$ in both simulation and real-world environments. In simulation, Habilis-$β$ achieves 572.6 TPH and 39.2 s MTBI (vs. 120.5 TPH and 30.5 s for $π_{0.5}$), while in a real-world humanoid logistics workflow it achieves 124 TPH and 137.4 s MTBI (vs. 19 TPH and 46.1 s for $π_{0.5}$). Finally, Habilis-$β$ achieves the highest reported performance on the standard RoboTwin 2.0 leaderboard across representative tasks, validating its effectiveness in complex manipulation scenarios.

Problem

• vision-language-action
• fast-motion
• long-lasting
• on-device
• continuous-run evaluation

Innovation

• vision-language-action
• on-device inference
• continuous-run evaluation
• rectified-flow distillation
• classifier-free guidance