← Leaderboard
LaST-R1
VLACUHK · Apr 30, 2026
VLA trained with RL post-training after a one-demo-per-task SFT warm-up; it holds the top LIBERO score under that non-standard data regime.
- Type
- VLA
- Parameters
- 4B
- Backbone
- Qwen3-VL-4B
- Weights
- Open
- License
- Apache-2.0
- Commercial use
- Allowed
- Released
- Apr 30, 2026
- Robots
- Franka Panda (sim), Franka Research 3
- SO-101
- –
- LeRobot
- –
Benchmark results
1 of 13 benchmarks- Paper / model card ↗Apr 2026· n = 2000
#1 of 4699.9%, self-reported by the authors, not a leaderboard entry
Not reported on 12 benchmarks
Adoption
updated Nov 17, 2026Runs on
Wrong or missing number? Report an error