X-Planner: Event-Structured Task Planning for Embodied Intelligence
X-Planner adds event-structured discrete and latent plans for long-horizon vision-language-action robot tasks.
X-Planner is a planning front-end for long-horizon embodied manipulation that makes intermediate task structure explicit for vision-language-action systems. Its data combine Ego, UMI, and teleoperation with source-dependent annotation depth, takeover-time labels, and human-designed failures. A shared VLM exposes discrete event states and continuous chain-of-thought states through Staircase Decoding, anchored by frozen latent-to-text reconstruction. Offline, it placed second of four models on BERTScore-F1 and a judge Overall score, and real-robot runs beat the evaluated baselines.
- Mixes Ego, UMI, and teleoperation data with hierarchy-aware annotations.
- Shared VLM emits discrete events or latent states via Staircase Decoding.
- Ranked second of four models on BERTScore-F1 and judge Overall.
- Real-robot experiments outperformed the evaluated baselines.
Full article168 words · extracted from huggingface.co · click to collapse
Task planning bridges high-level instructions and executable behavior in long-horizon manipulation, yet modern Vision-Language-Action (VLA) systems often leave this intermediate structure implicit. Existing chain-of-thought (CoT) planners also tend to rely on coarse task-level annotations or serialize long reasoning traces token by token. We present X-Planner, a planning front-end that addresses both the supervision and representation of embodied reasoning. Our planning data combine Ego, UMI, and teleoperation under a hierarchy granularity with source-dependent annotation depth. Takeover-time annotations and human-designed failures supervise ongoing error recognition. On the model side, a shared VLM backbone exposes two event-structured plan forms: a discrete interface that emits interpretable event states and a latent interface that relays continuous CoT states across staggered Transformer depths through Staircase Decoding. A frozen latent-to-text reconstruction objective provides a semantic anchor for the latent representation. Offline two-step planning evaluation places X-Planner second among four evaluated models on both BERTScore-F1 and a judge-based Overall score. In real-robot experiments, respectively, outperforming the evaluated baselines. These results characterize planning-text quality and downstream execution.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.25187