List your company

Reasoning on Blocks World

100Success RateWith parent pointer
89.18491.99294.897.608May 29, 2026

Evaluation results

MethodLinks
With parent pointerTraining=GRPO2026.051007.31
BFSTraining=RL2026.0599.88.56
BFSTraining=pretrained2026.0599.79.54
No parent pointerTraining=GRPO2026.0597.38.25
No parent pointerTraining=SFT2026.05909.12
With parent pointerTraining=SFT2026.0589.69.06