Unverified50% confidenceFactExact time
A per-step success rate that looks very high gets rapidly diluted across a long chain, with degradation becoming more severe as the number of steps increases.
1
Sources
50%
Confidence
Long-term
Relevance
10/7/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSparse rewards combined with long time spans cause gradient-based optimization signals to decay significantly during backpropagation.76% similarUnverifiedIn multi-step tasks, tool-call failures accumulate, and a high single-step failure rate causes the overall success rate to decline exponentially.76% similarVerifiedThe learning rate determines how large each step is: too large causes overshooting the valley bottom and divergence, while too small makes training extremely time-consuming.73% similarUnverified扩散模型步数越多理论上结果越精细，但耗时也成比例增长，20步是速度与质量的经典平衡点72% similarUnverifiedWhen the number of steps drops sharply to 4, the denoising span required per step increases significantly, forcing the model to make bolder predictions in a single step, which can easily lead to loss of high-frequency detail and incoherent motion trajectories.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/985220API
curl https://kongchang.com/api/v1/knowledge/claims/985220MCP
get_claim(id=985220)