MOTHRAG cuts multi-hop search work without a knowledge graph
MOTHRAG is an for in . It avoids building a before search. Systems such as GraphRAG, HippoRAG, and RAPTOR can score well, but they depend on an offline graph-building step. When the data changes, that graph often needs a heavy run again.
This becomes expensive for data that changes daily, such as prices, company filings, , or news. MOTHRAG instead uses a graph-free , then coordinates the search steps at query time. New data can be embedded and appended without rebuilding the graph or retraining. The reported benchmark scores were 78.1 on HotpotQA, 76.3 on 2WikiMultiHopQA, and 50.5 on MuSiQue, ahead of GraphRAG, HippoRAG, and RAPTOR in the shared table.
The stated cost is about $0.03 per query using ordinary APIs and no GPU. It is not a clear win against GPU-based systems such as NeocorRAG, where the HotpotQA result was roughly tied at 78.1 versus 78.
Key points
- MOTHRAG handles without a .
- Updates are described as embed-and-append, with no graph rebuild or retraining step.
- It beat GraphRAG, HippoRAG, and RAPTOR in the reported HotpotQA, 2WikiMultiHopQA, and MuSiQue scores.
- The stated cost is about $0.03 per query on ordinary APIs, with no GPU required.
- It only roughly matched, rather than clearly beat, GPU-based NeocorRAG on HotpotQA.