Unverified50% confidenceFactExact time
Embedding models are trained to bring texts with similar meaning close together, so they smooth out character sequences that carry no semantics, such as random product IDs, making it hard for semantic search to match such queries exactly.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Building a Precise AI Agent Knowledge Base: Hybrid Retrieval and Reranking in Practice
bilibili双语对照翻译10/19/2026
Related Entities
Related Claims
Unverified模型的格式方言源于指令微调阶段训练数据的格式差异，被模型深度内化形成强烈格式偏好，难以通过运行时提示词完全覆盖76% similarUnverifiedThe capability boundaries of intent recognition systems are determined by manually written rules and have limited scalability, whereas large language models, through pretraining on massive text, can handle new tasks that were not explicitly programmed, achieving a paradigm shift from rule-driven to data-driven approaches.74% similarUnverifiedClassifiers are highly sensitive to the phrasing of requests; asking the model directly to find vulnerabilities is easily intercepted and routed, whereas describing the same task using more neutral programming-language framing often does not trigger restrictions.73% similarUnverifiedMainstream text embedding models are typically trained based on contrastive learning, mapping semantically similar text pairs to nearby positions in vector space and pushing unrelated text pairs further apart.73% similarUnverifiedPrompt injection exploits not a vulnerability in the code parser, but the model's own insufficient awareness of the boundary between instructions and data72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1071362API
curl https://kongchang.com/api/v1/knowledge/claims/1071362MCP
get_claim(id=1071362)