Unverified50% confidenceSolutionExact time
In reward function design, sparse rewards can easily cause training to stagnate, and introducing dense reward signals such as distance reduction can accelerate convergence.
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSparse rewards combined with long time spans cause gradient-based optimization signals to decay significantly during backpropagation.77% similarUnverifiedWhen a model is optimized to maximize a reward signal, any behavior that improves the score may be learned; researchers refer to this as reward hacking or specification gaming75% similarUnverified研究者提出长度归一化奖励方法，即在计算奖励分数时除以回答长度，使模型无法通过单纯增加长度获取更高奖励74% similarUnverifiedIf the reward shaping signal is too weak, the AI will take shortcuts; if it is too strong or poorly designed, it may introduce new exploits or constrain the exploration space.73% similarUnverifiedEntropy collapse is especially prominent in long Chain-of-Thought training, because reward signals are only given at the end of the reasoning chain and gradient signals at intermediate steps are extremely sparse.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/990236API
curl https://kongchang.com/api/v1/knowledge/claims/990236MCP
get_claim(id=990236)