MPF: Aligning and Debiasing Language Models post Deployment via Multi Perspective Fusion
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908431912271872 |
|---|---|
| author | Guan, Xin Lin, PeiHsin Wu, Zekun Wang, Ze Zhang, Ruibo Kazim, Emre Koshiyama, Adriano |
| author_facet | Guan, Xin Lin, PeiHsin Wu, Zekun Wang, Ze Zhang, Ruibo Kazim, Emre Koshiyama, Adriano |
| contents | Multiperspective Fusion (MPF) is a novel posttraining alignment framework for large language models (LLMs) developed in response to the growing need for easy bias mitigation. Built on top of the SAGED pipeline, an automated system for constructing bias benchmarks and extracting interpretable baseline distributions, MPF leverages multiperspective generations to expose and align biases in LLM outputs with nuanced, humanlike baselines. By decomposing baseline, such as sentiment distributions from HR professionals, into interpretable perspective components, MPF guides generation through sampling and balancing of responses, weighted by the probabilities obtained in the decomposition. Empirically, we demonstrate its ability to align LLM sentiment distributions with both counterfactual baselines (absolute equality) and the HR baseline (biased for Top Univeristy), resulting in small KL divergence, reduction of calibration error and generalization to unseen questions. This shows that MPF offers a scalable and interpretable method for alignment and bias mitigation, compatible with deployed LLMs and requiring no extensive prompt engineering or finetuning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_02595 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | MPF: Aligning and Debiasing Language Models post Deployment via Multi Perspective Fusion Guan, Xin Lin, PeiHsin Wu, Zekun Wang, Ze Zhang, Ruibo Kazim, Emre Koshiyama, Adriano Computation and Language Artificial Intelligence Multiperspective Fusion (MPF) is a novel posttraining alignment framework for large language models (LLMs) developed in response to the growing need for easy bias mitigation. Built on top of the SAGED pipeline, an automated system for constructing bias benchmarks and extracting interpretable baseline distributions, MPF leverages multiperspective generations to expose and align biases in LLM outputs with nuanced, humanlike baselines. By decomposing baseline, such as sentiment distributions from HR professionals, into interpretable perspective components, MPF guides generation through sampling and balancing of responses, weighted by the probabilities obtained in the decomposition. Empirically, we demonstrate its ability to align LLM sentiment distributions with both counterfactual baselines (absolute equality) and the HR baseline (biased for Top Univeristy), resulting in small KL divergence, reduction of calibration error and generalization to unseen questions. This shows that MPF offers a scalable and interpretable method for alignment and bias mitigation, compatible with deployed LLMs and requiring no extensive prompt engineering or finetuning. |
| title | MPF: Aligning and Debiasing Language Models post Deployment via Multi Perspective Fusion |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2507.02595 |