Verified70% confidenceFactExact time
Large language models have hallucination defects in code generation scenarios, manifested as referencing non-existent library functions and generating syntactically correct but logically erroneous code.
4
Sources
70%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
UnverifiedIn code generation scenarios, hallucination manifests as calling non-existent API methods, generating syntactically correct but logically erroneous code, or incorrectly referencing deprecated library versions.83% similarUnverifiedIn programming scenarios, models may fabricate non-existent API functions, invent version numbers for libraries, or produce code solutions with logical flaws.80% similarUnverified实验数据显示，语法约束解码使小型语言模型生成Bash代码的语法正确率大幅提高，引号不匹配和结构不完整等错误几乎被完全消除77% similarUnverifiedLarge models perform better when processing decompiled pseudocode (such as C-like pseudocode output by Ghidra), because the training corpus contains a large amount of high-level language code, while their ability to understand raw assembly instructions is relatively limited.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577951API
curl https://kongchang.com/api/v1/knowledge/claims/577951MCP
get_claim(id=577951)