Unverified50% confidenceBenchmarkExact time
The plain diff method achieved a label-aware F1 of 0.525 and a localization F1 of 0.826.
1
Sources
50%
Confidence
Long-term
Relevance
9/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDiff requires a strict format (such as unified diff, line matching), and any single deviation can cause the entire patch to fail to apply.62% similarUnverifiedDiff improvement schemes targeting text provenance include Patience diff, semantic diff, and block-move detection algorithms.58% similarUnverifiedUsing a simple text similarity threshold or difference detection can easily misclassify reasonable wording changes as violations.57% similarUnverified当检索到的文本块余弦相似度低于0.60时，图会完全绕过LLM合成环节，转而向用户请求澄清或回退到实时网络搜索57% similarUnverified向量化通常将文本映射为 768 维或 1536 维的浮点数向量，相似度检索采用余弦相似度或近似最近邻算法如 HNSW、IVF-PQ56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/919720API
curl https://kongchang.com/api/v1/knowledge/claims/919720MCP
get_claim(id=919720)