Mitigating Source Bias with LLM Alignment
Recent studies have revealed a phenomenon known as source bias, where PLM-based retrievers assign higher relevance scores to LLM-generated content despite its semantic quality being comparable to human-written content. As LLMs rapidly advance and become more widely used, effectively counteracting source bias is crucial for the sustainable development of the information retrieval (IR) ecosystem. Existing methods primarily attempt to address source bias from the retriever side, adopting a "passive defense" approach that intervenes only after biased content has entered the retrieval pipeline. These solutions are limited by frequent retriever updates in industrial applications, high recurring costs, and their inability to address the root cause of source bias.
Authors
- Gang Wang (ORCID: https://orcid.org/0000-0002-8795-8953)
- Liang Pang (ORCID: https://orcid.org/0000-0003-1161-8546)
- Jun Xu (ORCID: https://orcid.org/0000-0001-7170-111X)
- Zhaocheng Du (ORCID: https://orcid.org/0000-0002-1811-129X)
- Sunhao Dai (ORCID: https://orcid.org/0009-0002-7549-0860)
- Yuqi Zhou (ORCID: https://orcid.org/0009-0008-2453-9138)
- Zhuoyang Li (ORCID: https://orcid.org/0000-0002-2248-2982)
Institutions
- Chinese Academy of Sciences (CN)
- Huawei Technologies (China) (CN)
- Institute of Computing Technology (CN)
- Renmin University of China (CN)
Publication Details
- Published
- 2025-07-13
- DOI
- https://doi.org/10.1145/3726302.3730038
- Citations
- 1
- Primary Topic
- Topic Modeling
- Type
- article
- Field-Weighted Citation Impact
- 2.48
Funders
- National Natural Science Foundation of China
- Renmin University of China
- Youth Innovation Promotion Association of the Chinese Academy of Sciences
- Fundamental Research Funds for the Central Universities
- Youth Innovation Promotion Association