Unverified50% confidenceBenchmarkExact time
In the experiment of writing tok from only a one-page reference document, the strongest model used 25% and 19% fewer tokens to write tok than to write BPMN XML.
1
Sources
50%
Confidence
Long-term
Relevance
11/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFour language models that had never previously encountered tok, relying only on a one-page reference document, achieved valid tok generation rates of 85.7% to 100% on 87 process descriptions, while the corresponding BPMN XML could be generated for every description.74% similarUnverifiedTokenPrune has not yet been validated in large-scale production environments; the 70%–90% reduction figures come from the author's own tests, and actual results will vary depending on the type of API workload.66% similarUnverified系统用轻量小模型对每篇论文摘要进行 1-10 分的相关性评分，评分依据是用户编写的描述研究兴趣的 Markdown 文件65% similarUnverifiedWhen small models are made to do verbatim quotations, only 24% to 35% of the quotations pass verification against the source text.65% similarUnverifiedThe model offers three reasoning levels: low, high and max; in the author's tests max generated about 16% more tokens than high, with almost the same total time for reasoning tasks.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1142421API
curl https://kongchang.com/api/v1/knowledge/claims/1142421MCP
get_claim(id=1142421)