Unverified50% confidenceFactExact time
DPR用双塔神经网络分别编码问题和文档，通过内积衡量相关性，相较传统BM25能捕捉更深层的语义关联
1
Sources
50%
Confidence
Long-term
Relevance
7/25/2026
First Seen
Sources
RAG检索增强生成详解：原理、工作流程与核心优势
bilibiliAI小公举-timi7/9/2026
Related Claims
Unverified对比学习在文本检索领域的突破来自2021年的DPR和SimCSE；DPR验证了双塔结构+批内负例策略在开放域问答检索上可超越BM2567% similarUnverifiedA dense dual-encoder consists of two independent neural networks—a query encoder and a document encoder—that map queries and documents into the same vector space.65% similarUnverifiedBM25引入词频饱和和文档长度归一化两个关键饱和函数62% similarUnverifiedThe root cause of prompt injection lies in the fact that LLM architecture encodes system prompts, user inputs, and externally retrieved content uniformly into the same continuous token sequence, making it impossible for the model to distinguish trusted instructions from untrusted data at the semantic level.62% similarUnverifiedGlobal self-attention has a modeling complexity of O(n²), and subsequent research has developed relative positional encoding schemes such as RoPE and ALiBi to enhance long-text processing capability.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/615566API
curl https://kongchang.com/api/v1/knowledge/claims/615566MCP
get_claim(id=615566)