Research map / Large language models / Retrieval and knowledge
Towards Whole-corpus Reconstruction of Heterogeneous RAG Knowledge Bases
Peiru Yang, Yi Luo, Zhenfeng Gao, Tong Ju, Haoran Zheng, Linjie Zhu, Hongke Fu, Qing Li, Shangguang Wang, Tao Qi
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 |
|---|---|---|---|
| use GeoEx embedding-space planning with orthogonal jumps and local perturbation instead of local text-driven query expansion | RAG extraction attacks · heterogeneous 4-domain corpus | coverage: avg +19.94% coverage over best baseline | Improved |
| use a surrogate encoder aligned with the target retriever (A) instead of misaligned (M) (ablation) | GeoEx · Contriever, avg over 5 LLMs | coverage: 44.21 -> 49.35 | Improved |
| remove global orthogonal exploration (ablation) | GeoEx · ablation across retrievers | coverage | Improved |
| add local Gaussian densification (ablation) | GeoEx · ablation across retrievers | coverage | Improved |
| use orthogonal projection instead of high-magnitude isotropic noise for global jumps (ablation) | GeoEx · ablation across retrievers | coverage | Improved |
| increase inverse decoding length from 120 to beyond (ablation) | GeoEx · Contriever, 4 domains | coverage: 49.80 -> 32.40 at longest length | Hurt |
| increase number of domains from 4 to 8 (ablation) | GeoEx · heterogeneous corpus | coverage | Hurt |
| increase number of domains from 4 to 8 (ablation) | GeoEx · heterogeneous corpus | efficiency: 70.17 -> 83.67 | Improved |
| swap surrogate encoder among seven embedding models (ablation) | GeoEx · coverage curves | coverage | No clear effect |
| increase number of orthogonal targets Ng (ablation) | GeoEx · sensitivity analysis | coverage | Mixed |
| apply reranking, prompt rewriting or threshold filtering defenses | GeoEx · Contriever/BGE/E5 with GPT-4o | coverage | No clear effect |
| apply DGEA with misaligned surrogate encoder | DGEA · Contriever, avg over 5 LLMs | coverage: 20.10 -> 12.71 | Hurt |
Closest papers by meaning
- Follow My Instruction and Spill the Beans: Scalable Data Extraction from Retrieval-Augmented Generation Systems (ICLR 2025 · similarity 0.88)
- SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG (NeurIPS 2025 · similarity 0.87)
- From Complex to Atomic: Enhancing Augmented Generation via Knowledge-Aware Dual Rewriting and Reasoning (ICML 2025 · similarity 0.87)
- Can Knowledge-Graph-based Retrieval Augmented Generation Really Retrieve What You Need? (NeurIPS 2025 · similarity 0.87)
- GFM-RAG: Graph Foundation Model for Retrieval Augmented Generation (NeurIPS 2025 · similarity 0.86)
- RAGGED: Towards Informed Design of Scalable and Stable RAG Systems (ICML 2025 · similarity 0.86)
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.
