Vision-Predictor-Action: Towards Efficient Embodied AI via Neuro-Symbolic JEPA Predictors
Noch nicht auf Deutsch verfügbar: Anzeige im englischen Original.
Aus dem Abstract
Vision-Language-Action (VLA) models commonly generate robot actions token by token through Large Language Model (LLM) decoders, causing inference latency to grow with action sequence length and limiting closed-loop control frequency. We present Goal-Conditioned Vision-Predictor-Action (VPA), which replaces autoregressive decoders with a lightweight latent control loop.
Aus dem Abstract. Unsere Zusammenfassung folgt.