Unverified50% confidenceTradeoffExact time
BLEU is insensitive to synonym substitution and word order variation, and its correlation with human judgment is particularly weak for morphologically rich languages.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedText similarity metrics such as BLEU and ROUGE are difficult to use for evaluating the quality of legal headnotes, because legally correct formulations may use entirely different wording from the reference answers.75% similarVerifiedText generated by large language models tends to have low perplexity, with word choices that are overly consistent with probabilistic expectations, lacking the unexpectedness of human writing.71% similarUnverifiedBM25 has no understanding of semantics whatsoever, matches word by word, and cannot handle synonym substitution or context-dependent relationships.70% similarUnverified热词偏置的偏置权重设置过高会导致误召回，设置过低则无法有效纠正音译行为70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949203API
curl https://kongchang.com/api/v1/knowledge/claims/949203MCP
get_claim(id=949203)