Add robotics pipeline tag and remove arxiv tag from metadata

README.md

@@ -1,13 +1,13 @@
 ---
+datasets:
+- snupilab/humanoidtoolbench-teleop
 language:
 - en
 tags:
 - robotics
 - humanoidtoolbench
 - mujoco
-- arxiv:2610.02089
-datasets:
-- snupilab/humanoidtoolbench-teleop
+pipeline_tag: robotics
 ---
 
 # HumanoidToolBench DP: Simulation training, 3,003 segments
@@ -39,4 +39,4 @@ This is the DP policy reported in the HumanoidToolBench paper; its success rates
 
 Retain run/checkpoints/ckpt_40000 layout. Use HumanoidToolBench native adapter with run directory run and step 40000. Run from exported run directory for explicit dataset_statistics.json resolution; serialized normalization bounds are retained. Nearby launch.json contains required exact language-conditioning receipt. Install pinned Psi0 source plus HumanoidToolBench adapter and pre-cache openai/clip-vit-large-patch14 at 32bd64288804d66eefd0ccbe215aa642df71cc41. DP native serving selects ema_net.pth; model.safetensors is retained for launcher discovery. Do not substitute non-EMA weights.
 
-Training uses independent model optimizers and shared GPU execution through MPS. Publication is performed by a CPU uploader after final checkpoint validation.
+Training uses independent model optimizers and shared GPU execution through MPS. Publication is performed by a CPU uploader after final checkpoint validation.