Unverified50% confidenceFactExact time
Hinton在原始论文中引入温度参数T来控制概率分布的柔软程度，T越高分布越平滑，类间相似关系暴露越充分，这些类间关系被称为暗知识（Dark Knowledge）
1
Sources
50%
Confidence
Long-term
Relevance
8/18/2026
First Seen
Sources
Related Claims
UnverifiedThe higher the temperature parameter, the flatter the probability distribution output by Softmax, and the more creative and diverse the output; the lower the temperature, the sharper the distribution and the more deterministic and conservative the output.60% similarVerifiedThe non-determinism of LLMs stems from sampling strategies; the Temperature parameter controls the randomness of the output distribution, while Top-p and Top-k further constrain the range of candidate tokens.57% similarUnverified不确定性估计主要有两条技术路径：利用Token概率（logprobs）和蒙特卡洛采样（对同一输入进行5-20次高温度采样并计算香农熵）57% similarUnverified蒸馏的核心洞察在于神经网络的软标签（类别概率分布）包含比one-hot硬标签更丰富的结构信息56% similarUnverifiedTemperature controls the flatness of the sampling distribution; a higher value tends toward selecting lower-probability words, while Top-p dynamically narrows the candidate vocabulary set through a cumulative probability threshold.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/766985API
curl https://kongchang.com/api/v1/knowledge/claims/766985MCP
get_claim(id=766985)