Beyond Prompting: An Efficient Embedding Framework for Open-Domain Question Answering
Zhanghao Hu, Hanqi Yan, Qinglin Zhu, Zhenyi Shen, Yulan He, Lin Gui · 2025
Read Beyond Prompting: An Efficient Embedding Framework for Open-Domain Question Answering by Zhanghao Hu, Hanqi Yan, Qinglin Zhu, Zhenyi Shen, Yulan He, Lin Gui (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-01606.txt.
Description
Zhanghao Hu and co-authors present an embedding-level framework for open-domain question answering. Their 2025 system, called EmbQA, aims to cut the cost and wobble of prompt-heavy pipelines. Such pipelines lean on many rounds of instructions to the model.
Query vectors are refined by light linear layers trained without labels, by a contrastive goal. That step re-ranks found passages so answer-bearing ones rise. A second trick widens the model's inner space to spawn varied candidate answers. An entropy rule then picks the most confident one. Tests span three models, three retrievers, and four benchmarks. Gains in both accuracy and speed over recent rivals are reported.
Catalog record
| CREATOR | Zhanghao Hu, Hanqi Yan, Qinglin Zhu, Zhenyi Shen, Yulan He, Lin Gui |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-01606