Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• CompactRAG reduces LLM calls to only two invocations during inference regardless of the number of reasoning hops.method 95%
• CompactRAG achieves competitive accuracy on HotpotQA, 2WikiMultiHopQA, and MuSiQue datasets compared to iterative RAG baselines.result 90%
• CompactRAG substantially reduces token consumption compared to iterative RAG baselines.result 90%
• CompactRAG's offline stage converts corpus into an atomic QA knowledge base with minimal, fine-grained question-answer pairs.method 85%
• CompactRAG's online stage uses RoBERTa-based answer extraction instead of LLM for intermediate reasoning steps.technical 80%
• CompactRAG's efficiency depends on the quality of the initial corpus transformation, which can be computationally intensive upfront.limitation 85%
• CompactRAG offers a cost-efficient and practical approach to multi-hop reasoning over large knowledge corpora.market 80%
• CompactRAG's success in sub-question decomposition accuracy could vary depending on the complexity of input questions.limitation 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card