Knowledge Compression via Question Generation: Enhancing Multihop Document Retrieval without Fine-tuning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913896344846336 |
|---|---|
| author | Eponon, Anvi Alex Shahiki-Tash, Moein Batyrshin, Ildar Maldonado-Sifuentes, Christian E. Sidorov, Grigori Gelbukh, Alexander |
| author_facet | Eponon, Anvi Alex Shahiki-Tash, Moein Batyrshin, Ildar Maldonado-Sifuentes, Christian E. Sidorov, Grigori Gelbukh, Alexander |
| contents | This study presents a question-based knowledge encoding approach that improves retrieval-augmented generation (RAG) systems without requiring fine-tuning or traditional chunking. We encode textual content using generated questions that span the lexical and semantic space, creating targeted retrieval cues combined with a custom syntactic reranking method.
In single-hop retrieval over 109 scientific papers, our approach achieves a Recall@3 of 0.84, outperforming traditional chunking methods by 60 percent. We also introduce "paper-cards", concise paper summaries under 300 characters, which enhance BM25 retrieval, increasing MRR@3 from 0.56 to 0.85 on simplified technical queries.
For multihop tasks, our reranking method reaches an F1 score of 0.52 with LLaMA2-Chat-7B on the LongBench 2WikiMultihopQA dataset, surpassing chunking and fine-tuned baselines which score 0.328 and 0.412 respectively.
This method eliminates fine-tuning requirements, reduces retrieval latency, enables intuitive question-driven knowledge access, and decreases vector storage demands by 80%, positioning it as a scalable and efficient RAG alternative. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_13778 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Knowledge Compression via Question Generation: Enhancing Multihop Document Retrieval without Fine-tuning Eponon, Anvi Alex Shahiki-Tash, Moein Batyrshin, Ildar Maldonado-Sifuentes, Christian E. Sidorov, Grigori Gelbukh, Alexander Information Retrieval Artificial Intelligence Computation and Language This study presents a question-based knowledge encoding approach that improves retrieval-augmented generation (RAG) systems without requiring fine-tuning or traditional chunking. We encode textual content using generated questions that span the lexical and semantic space, creating targeted retrieval cues combined with a custom syntactic reranking method. In single-hop retrieval over 109 scientific papers, our approach achieves a Recall@3 of 0.84, outperforming traditional chunking methods by 60 percent. We also introduce "paper-cards", concise paper summaries under 300 characters, which enhance BM25 retrieval, increasing MRR@3 from 0.56 to 0.85 on simplified technical queries. For multihop tasks, our reranking method reaches an F1 score of 0.52 with LLaMA2-Chat-7B on the LongBench 2WikiMultihopQA dataset, surpassing chunking and fine-tuned baselines which score 0.328 and 0.412 respectively. This method eliminates fine-tuning requirements, reduces retrieval latency, enables intuitive question-driven knowledge access, and decreases vector storage demands by 80%, positioning it as a scalable and efficient RAG alternative. |
| title | Knowledge Compression via Question Generation: Enhancing Multihop Document Retrieval without Fine-tuning |
| topic | Information Retrieval Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2506.13778 |