Unverified50% confidenceFactExact time
自回归生成机制下，生成长度越长累积计算量越大，生成1000个Token的耗时大约是生成100个Token的10倍
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
Unverified一个能力提升10%但速度翻倍的模型，实际生产力的增幅可能远不止10%65% similarUnverifiedModel performance is most stable when the context is kept within 100K tokens, and the hallucination rate rises noticeably beyond 100K.64% similarUnverified将一个100步任务拆为10个10步子任务，理论上可将总Token消耗降低近90%60% similarUnverified蒙特卡洛方法的收敛速度为O(1/√N)，精度每提高一个数量级需要样本量增加百倍59% similarUnverifiedResearch indicates that for every one-second increase in initial load time, the user drop-off rate increases by approximately 10%.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/582550API
curl https://kongchang.com/api/v1/knowledge/claims/582550MCP
get_claim(id=582550)