X-Policy (4B)
This repository contains the X-Policy inference checkpoint used for the formal VLA-Arena leaderboard evaluation.
Checkpoint
- Model: X-Policy
- Size: 4B parameters
- Selected training step: 18,000
- Published weights: inference parameters
X-Policy improves closed-loop temporal memory, task-routed action modeling, and replanning consistency.
The repository contains params/, assets/, and _CHECKPOINT_METADATA.
Optimizer state and raw training state are intentionally excluded.
VLA-Arena evaluation
X-Policy was evaluated on all 33 VLA-Arena suite/level cells using a fixed replanning interval of five environment steps.
- Episodes: 1,700
- Successes: 924
- Cell-mean success rate: 53.21%
- Task-weighted success rate: 54.35%
- L0 / L1 / L2 cell-mean SR: 76.18% / 50.18% / 33.27%
The task-level scores and episode counts are available in
evaluation/results.json.
Loading note
This is an Orbax checkpoint for X-Policy. It is not a Transformers
from_pretrained checkpoint.