Unverified50% confidenceFactExact time
Top-1精度衡量量化模型预测的最高概率token与原始模型预测的最高概率token相匹配的比例
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
Related Claims
UnverifiedTop-1% accuracy is a metric for evaluating the output consistency between a quantized model and the original full-precision model, focusing on whether the top 1% of tokens with the highest model confidence are consistent with the full-precision version.74% similarVerifiedTop-P (nucleus sampling) samples from the smallest set of tokens whose cumulative probability exceeds the threshold P, while Top-K selects only from the K candidates with the highest probabilities.67% similarUnverifiedGLM 5.2 scored 51 points on the Artificial Analysis Intelligence Index in its MAX version, making it the highest-scoring open-weight model currently available.59% similarUnverified在MoE模型中每个token只激活少量Expert（通常2-4个，Top-K路由），实际计算量可能仅相当于6-7B的Dense模型58% similarUnverified2.4万亿参数模型在BF16精度下考虑KV Cache和激活值后实际约需1.4-1.5TB显存57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/705394API
curl https://kongchang.com/api/v1/knowledge/claims/705394MCP
get_claim(id=705394)