Expired50% confidenceBenchmarkExact time
The coding capabilities of top large language models have already surpassed most human programmers on benchmarks such as HumanEval.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/20/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/18/2026(expired)
Sources
Related Claims
UnverifiedThe structural weakness of large language models is that they optimize for 'code that looks correct' rather than 'code that meets the requirements of the current stage,' and they replicate high-frequency patterns from training data.71% similarUnverifiedCommon benchmarks for evaluating large language model capabilities include MMLU, HumanEval, MATH, and LMSYS Chatbot Arena.70% similarUnverified使用领域专业术语的Prompt相比自然语言描述，能使大语言模型的代码生成准确率提升40%以上67% similarUnverified当前自然语言生成代码的可靠性仍有明显瓶颈，复杂业务逻辑场景下生成的代码往往需要人工审查和修改66% similarUnverifiedPerplexity measures a language model's ability to predict test text; the lower the value, the less surprised the model is by the text.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/574787API
curl https://kongchang.com/api/v1/knowledge/claims/574787MCP
get_claim(id=574787)