Conceptreward sparsity / 稀疏奖励
Reward Sparsity
A common problem in reinforcement learning in which an agent receives reward signals only after completing a large number of steps, making credit assignment difficult and reducing training efficiency.
Timeline (last 90 days)
Sep 29
In sparse reward scenarios, an agent may complete thousands of steps before receiving a single signal, making credit assignment extremely difficult.
Unverified50%