MJ22x/

go-vla-benchmark

Fine-tuned VLA (Vision-Language-Action) models on a simulated 5×5 Go stone placement task using a Panda robot arm in robosuite/MuJoCo.

Each episode: pick up a stone from a source tray and place it at a specific intersection on a 5×5 Go board. The instruction specifies the stone color and target position (e.g., "Place a black stone on the Go board at row 3, column 1.").

	Models




	

Model Base Params… See the full description on the dataset page: https://huggingface.co/datasets/MJ22x/go-vla-benchmark.