Research map / Large language models / Retrieval and knowledge
Predictive Prefetching for Retrieval-Augmented Generation
Wuyang Zhang, Shichao Pei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add predictive asynchronous prefetching | Sync-RAG · Four QA benchmarks | E2E latency: 9.2s -> 5.2s | Improved |
| add predictive asynchronous prefetching | Sync-RAG · HotpotQA | EM: 69.2 -> 68.7 | Mixed |
| use learned prediction instead of stale-query pipeline | PipeRAG · HotpotQA | EM: 66.8 -> 68.7 | Improved |
| add asynchronous architecture (ablation) | Full system · HotpotQA | TTFT: 287ms -> 108ms | Improved |
| add retrieval predictor (ablation) | Full system · HotpotQA | EM: 65.1 -> 68.7 | Improved |
| add online policy-gradient adaptation (ablation) | Full system · HotpotQA | EM: 66.2 -> 68.7 | Improved |
| add T5 query generator (ablation) | Full system · HotpotQA | EM: 67.1 -> 68.7 | Improved |
| add adaptive waiting (ablation) | Full system · HotpotQA | EM: 66.8 -> 68.7 | Improved |
| add sufficiency check (ablation) | Full system · HotpotQA | EM: 67.8 -> 68.7 | Improved |
| online adaptation on top of pretraining (ablation) | Retrieval Predictor · Code (RepoBench, OOD) | AUROC: 0.68 -> 0.77 | Improved |
| use transformer-internal signals instead of current entropy | Retrieval Predictor · entropy-crossing prediction | AUROC: 0.66 -> 0.81 | Improved |
| wait 3-4 tokens before issuing query (ablation) | Context Monitor · factual queries | query relevance score: 0.65 -> 0.86 | Improved |
Closest papers by meaning
- RAPID: Long-Context Inference with Retrieval-Augmented Speculative Decoding (ICML 2025 · similarity 0.88)
- Inference Scaling for Long-Context Retrieval Augmented Generation (ICLR 2025 · similarity 0.88)
- Accelerating Inference of Retrieval-Augmented Generation via Sparse Context Selection (ICLR 2025 · similarity 0.87)
- Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting (ICLR 2025 · similarity 0.87)
- RAGGED: Towards Informed Design of Scalable and Stable RAG Systems (ICML 2025 · similarity 0.87)
- A Theory for Token-Level Harmonization in Retrieval-Augmented Generation (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
