AlphaNIST: Emergent Assembly Skills from Binary Task Rewards

AlphaNIST achieves a 98.2% training success rate across 20 NIST Assembly Task Board #1 tasks using only sparse binary rewards. By training a single policy from scratch without demonstrations or task decomposition, the method demonstrates that reinforcement learning can discover assembly strategies through task diversity alone.