Unverified50% confidenceFactExact time
Temperature divides the logits by a coefficient before applying softmax, changing the steepness of the entire distribution; top-p first accumulates probabilities from high to low, truncating once the cumulative probability reaches p, removing low-probability tokens outside the long tail from the candidate set, then renormalizing and sampling within the remaining set.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTop-p（核采样）通过限定候选token的累积概率范围来控制输出多样性，与Temperature配合使用可实现更精细的输出调节75% similarVerifiedTemperature 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.74% similarUnverifiedThe sampling mentioned in decoding randomness includes temperature scaling, top-k and top-p, aimed at avoiding stiff, repetitive output, at the cost that low-probability tokens are occasionally selected and errors can be plausibly continued.69% 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.67% similarUnverified调低temperature参数并配置no_speech_threshold和logprob_threshold可降低模型在不确定片段时的幻觉66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1029978API
curl https://kongchang.com/api/v1/knowledge/claims/1029978MCP
get_claim(id=1029978)