RAGFlow 论文提出知识编译层：查询 token 减少 91%，FRAMES 延迟从 140 秒降到 30 秒
Context as Infrastructure: Knowledge Compilation for Retrieval in Agentic Systems with Harnesses
RAGFlow 这篇论文实测了知识编译到底有没有用：准确率没涨，但 token 省了 38%，延迟从 140 秒降到 30 秒，做 RAG 的可以看看怎么落地。
arXiv 论文提出把 RAG 中的检索循环升级为知识编译管道，用 LLM 为文档生成导航树、实体图谱、时间线等结构化产物。实验显示，仅凭 BM25/grep 加 MiniMax M3 的搜索 harness 就能在 BrowseComp-Plus 达到 85%、多跳基准 FRAMES 达到 91%，无需向量索引。编译产物不提升准确率，但作为供给层有效：claim-evidence 导航树平均减少 38% 查询 token，证据充足时削减 91%。决策门控将延迟从 140 秒降到 30 秒，RAGFlow 的开源实现将循环拆为五个规范阶段。
Context as Infrastructure: Knowledge Compilation for Retrieval in Agentic Systems with Harnesses
Retrieval-augmented generation is becoming agentic. A search harness writes queries, reads results, reformulates, and decides when to stop. This largely solves accuracy: a harness with only BM25/grep and MiniMax M3 answers 85% of BrowseComp-Plus and 91% on multi-hop FRAMES in our runs, without compiled or vector indexes. It exposes a context problem: an undisciplined loop repeats queries, rereads documents, and fills the window with unused passages. The missing layer is knowledge compilation: a configurable ingestion pipeline where LLMs enrich documents with context no chunk contains alone--navigation hierarchies with claim-evidence leaves, entity graphs, timelines, mind maps, ontologies, wikis, and memory. Compilation is not a better retriever: under a competent harness, structural artifacts gave no detectable accuracy gain on FRAMES, and a PageIndex-style outline degraded at scale. It is a supply layer. A navigation tree whose leaves are claim-evidence pairs--the mechanism we measure--cuts query tokens by 91% when evidence suffices and adds 1.9% when it does not, a 38% average reduction; this implies a 2% break-even sufficiency rate and a measured rate near 43%. A decision gate cuts latency from 140 s to 30 s, though substitution's token saving is bought with retrieval rounds and carries its own latency cost. Other artifacts--ontology, graph, wiki, timeline, mind map, memory--pay off in deployment (governance, integration, onboarding, temporal orientation, continuity) where accuracy benchmarks cannot see them. We describe RAGFlow's pipeline and harness (open source at https://github.com/infiniflow/ragflow), which turns the loop into five disciplined stations and keeps the context entering the model the smallest sufficient set; together they form an agent-native database: a store that hands the agent compiled, governed context instead of a ranked list of passages.
- Artificial Analysis10-28 21:38原文