Unverified50% confidenceFactExact time
温度参数（Temperature）的数学本质是对Softmax函数输出的概率分布进行缩放，趋近于0时模型进行贪心解码
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
AI Agent上线后如何做在线评估？四种方法全解析
Related Claims
VerifiedThe Temperature parameter acts on the softmax function output: when below 1, the probability distribution becomes sharper; when above 1, the distribution becomes flatter.85% similarUnverified温度参数通过对 softmax 层的 logits 进行缩放来控制概率分布的尖锐程度，温度趋近于0时接近贪心解码81% similarVerifiedThe temperature parameter scales the Logits vector, and when temperature is 0, it degenerates into greedy decoding.81% similarVerifiedThe large model temperature parameter adjusts the flatness of the output probability distribution by dividing logits by the temperature value before applying softmax normalization.80% similarUnverifiedThe Temperature parameter of large language models controls the randomness and creative divergence of the model; when Temperature is 0, the output is completely deterministic.80% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/510714API
curl https://kongchang.com/api/v1/knowledge/claims/510714MCP
get_claim(id=510714)