Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2506.17951 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911018793304064 |
|---|---|
| author | Tang, Quanwei Lee, Sophia Yat Mei Wu, Junshuang Zhang, Dong Li, Shoushan Cambria, Erik Zhou, Guodong |
| author_facet | Tang, Quanwei Lee, Sophia Yat Mei Wu, Junshuang Zhang, Dong Li, Shoushan Cambria, Erik Zhou, Guodong |
| contents | Recent advancements in retrieval-augmented generation (RAG) have enhanced large language models in question answering by integrating external knowledge. However, challenges persist in achieving global understanding and aligning responses with human ethical and quality preferences. To address these issues, we propose GraphMPA, a comprehensive graph-based framework with mode-seeking preference alignment. Our approach constructs a hierarchical document graph using a general similarity measurement, mimicking human cognitive processes for information understanding and synthesis. Additionally, we introduce mode-seeking preference optimization to better align model outputs with human preferences through probability-matching constraints. Extensive experiments on six datasets demonstrate the effectiveness of our \href{https://github.com/tangquanwei/GraphMPA}{GraphMPA}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_17951 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A Comprehensive Graph Framework for Question Answering with Mode-Seeking Preference Alignment Tang, Quanwei Lee, Sophia Yat Mei Wu, Junshuang Zhang, Dong Li, Shoushan Cambria, Erik Zhou, Guodong Computation and Language Recent advancements in retrieval-augmented generation (RAG) have enhanced large language models in question answering by integrating external knowledge. However, challenges persist in achieving global understanding and aligning responses with human ethical and quality preferences. To address these issues, we propose GraphMPA, a comprehensive graph-based framework with mode-seeking preference alignment. Our approach constructs a hierarchical document graph using a general similarity measurement, mimicking human cognitive processes for information understanding and synthesis. Additionally, we introduce mode-seeking preference optimization to better align model outputs with human preferences through probability-matching constraints. Extensive experiments on six datasets demonstrate the effectiveness of our \href{https://github.com/tangquanwei/GraphMPA}{GraphMPA}. |
| title | A Comprehensive Graph Framework for Question Answering with Mode-Seeking Preference Alignment |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2506.17951 |