← Leaderboard

NORA-1.5

VLA
Nanyang Technological University · Nov 18, 2025 · built on NORA

Successor to NORA that adds a flow-matching action expert to the autoregressive NORA backbone. It is post-trained with DPO on rewards from an action-conditioned world model (V-JEPA2-AC) and from ground-truth actions. Its LIBERO and SimplerEnv results (Google Robot only) are self-reported.

Type
VLA
Parameters
—
Backbone
Qwen2.5-VL-3B (NORA) + flow-matching action expert
Weights
Open
License
MIT
Commercial use
Allowed
Released
Nov 18, 2025
Robots
Franka Panda (sim), Google Robot (sim), Galaxea A1
SO-101
–
LeRobot
–
Inputs
RGB · L
VRAM
–

Benchmark results

1 of 13 benchmarks
LIBEROSaturatedSimulation
#52 of 7095.0%, self-reported by the authors, not a leaderboard entry
Not reported on 12 benchmarks
Wrong or missing number? Report an error