Context-Efficient Retrieval with Factual Decomposition
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_ | 1866917967908831232 |
|---|---|
| author | Li, Yanhong Yunis, David McAllester, David Zhou, Jiawei |
| author_facet | Li, Yanhong Yunis, David McAllester, David Zhou, Jiawei |
| contents | There has recently been considerable interest in incorporating information retrieval into large language models (LLMs). Retrieval from a dynamically expanding external corpus of text allows a model to incorporate current events and can be viewed as a form of episodic memory. Here we demonstrate that pre-processing the external corpus into semi-structured ''atomic facts'' makes retrieval more efficient. More specifically, we demonstrate that our particular form of atomic facts improves performance on various question answering tasks when the amount of retrieved text is limited. Limiting the amount of retrieval reduces the size of the context and improves inference efficiency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_19574 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Context-Efficient Retrieval with Factual Decomposition Li, Yanhong Yunis, David McAllester, David Zhou, Jiawei Computation and Language Information Retrieval There has recently been considerable interest in incorporating information retrieval into large language models (LLMs). Retrieval from a dynamically expanding external corpus of text allows a model to incorporate current events and can be viewed as a form of episodic memory. Here we demonstrate that pre-processing the external corpus into semi-structured ''atomic facts'' makes retrieval more efficient. More specifically, we demonstrate that our particular form of atomic facts improves performance on various question answering tasks when the amount of retrieved text is limited. Limiting the amount of retrieval reduces the size of the context and improves inference efficiency. |
| title | Context-Efficient Retrieval with Factual Decomposition |
| topic | Computation and Language Information Retrieval |
| url | https://arxiv.org/abs/2503.19574 |