Unverified50% confidenceFactExact time
The paper finds that fine-tuning on the query decomposition task exhibits better generalization than fine-tuning directly on the downstream RAGQ task.
1
Sources
50%
Confidence
Long-term
Relevance
11/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMulti-hop queries need to traverse multiple information nodes in a step-by-step manner, and vector retrieval is ill-suited for such scenarios due to its lack of a relational dimension.72% similarUnverifiedChunks that are too large dilute retrieval precision, while chunks that are too small result in incomplete semantics; the chunking strategy directly determines the retrieval quality of a RAG system.71% similarUnverifiedSome models with web retrieval access mix search results with their own generated content in their output, and if this is not distinguished when trusted, it easily creates a patchwork feel with a lack of argumentative connection between paragraphs.70% similarUnverifiedFine-tuning mainly changes a model's behavior on a specific narrow task; it can encode new information, but it is usually not the best choice for frequently changing or sensitive knowledge being cited, and retrieval or databases are usually more suitable.69% similarUnverifiedTaking RAG as an example, interviewers will follow up by asking why Fine-tuning was not used, how to optimize when retrieval recall rate is low, and which vector database was used and why.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1165027API
curl https://kongchang.com/api/v1/knowledge/claims/1165027MCP
get_claim(id=1165027)