Unverified50% confidenceFactExact time
通用Embedding模型（如OpenAI text-embedding-ada-002、BGE系列）在处理行业专有术语、产品代码或企业内部缩略语时表现欠佳，因为这些词汇在通用预训练语料中出现频率极低
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
企业级RAG知识库落地的5大工程硬伤：从Demo到生产的真实门槛
bilibili赋范课堂7/4/2026
Related Claims
UnverifiedLarge language models have two inherent weaknesses: their knowledge is cut off at training time and they cannot obtain the latest information, and they can only generate text and cannot directly perform actions.66% similarUnverifiedLLMs acquire language capabilities by performing unsupervised pre-training on massive text corpora (predicting the next token).66% similarUnverified对大语言模型而言，不要做X的约束有时比要做Y的指令更有效，因为模型在开放式创作任务中更容易被负样本锚定65% similarUnverifiedThe 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.65% similarUnverifiedThe safety alignment of many models degrades significantly in non-English languages, and low-resource languages are especially easy to exploit because the human annotation data for RLHF alignment training is highly concentrated in English.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/180992API
curl https://kongchang.com/api/v1/knowledge/claims/180992MCP
get_claim(id=180992)