Unverified50% confidenceFactExact time
LLM-generated text tends to select high-probability words and common collocations, causing the output to exhibit relatively low entropy in its vocabulary distribution.
1
Sources
50%
Confidence
Long-term
Relevance
9/5/2026
First Seen
Sources
Related Entities
Related Claims
Unverified研究表明，当 LLM 基于结构化输入（如 AST 提取的函数签名和调用关系）而非纯文本推测时，幻觉率会显著降低69% similarUnverifiedThe output of large language models is fundamentally probabilistic text generation, and a single rephrasing of a prompt can change field names, nesting levels, or enumeration values in the output without triggering compile-time or static analysis warnings.68% similarUnverifiedThe perplexity of LLM output is typically lower than that of human writing, and can be used to detect AI-generated content.67% similarUnverifiedLLMs learn the statistical patterns of word co-occurrence, grammatical structures, and rhetorical patterns in human language by training on massive text corpora, but the generation process does not involve understanding of content or intent.65% similarUnverifiedLLM的分词器（如BPE）为自然语言设计，会将数值拆解为多个不相关的token，破坏数字的连续性和大小语义65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/860442API
curl https://kongchang.com/api/v1/knowledge/claims/860442MCP
get_claim(id=860442)