Unverified50% confidenceBenchmarkExact time
Under pure vector retrieval with k=5, only 2 of 12 questions could assemble all the necessary evidence; traversing along the graph reached 11/12.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn a test of 12 questions about a 1,061-line pipeline file (three of which asked about features that do not exist), all three models scored a perfect 12/12 and did not invent any features.66% similarUnverifiedSelecting K models from N candidates is a combinatorial explosion problem; when N=20 and K=5, the number of combinations already exceeds 15,000.65% similarUnverifiedK折交叉验证的K通常取5或10，这在偏差-方差权衡上被实证为较优选择64% similarVerifiedMixtral 8x7B模型拥有8个专家，每个Token只激活2个，总参数量达到47B，但单次推理的计算量仅相当于一个13B参数的稠密模型63% similarUnverifiedIn a PR review comparison test, Opus found 23 valid issues (including 8 nits), Muse found 16, and Sol found only 9.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1062684API
curl https://kongchang.com/api/v1/knowledge/claims/1062684MCP
get_claim(id=1062684)