Assessing Bias in Metric Models for LLM Open-Ended Generation Bias Benchmarks
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917897810477056 |
|---|---|
| author | Demchak, Nathaniel Guan, Xin Wu, Zekun Xu, Ziyi Koshiyama, Adriano Kazim, Emre |
| author_facet | Demchak, Nathaniel Guan, Xin Wu, Zekun Xu, Ziyi Koshiyama, Adriano Kazim, Emre |
| contents | Open-generation bias benchmarks evaluate social biases in Large Language Models (LLMs) by analyzing their outputs. However, the classifiers used in analysis often have inherent biases, leading to unfair conclusions. This study examines such biases in open-generation benchmarks like BOLD and SAGED. Using the MGSD dataset, we conduct two experiments. The first uses counterfactuals to measure prediction variations across demographic groups by altering stereotype-related prefixes. The second applies explainability tools (SHAP) to validate that the observed biases stem from these counterfactuals. Results reveal unequal treatment of demographic descriptors, calling for more robust bias metric models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_11059 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Assessing Bias in Metric Models for LLM Open-Ended Generation Bias Benchmarks Demchak, Nathaniel Guan, Xin Wu, Zekun Xu, Ziyi Koshiyama, Adriano Kazim, Emre Computation and Language Artificial Intelligence Open-generation bias benchmarks evaluate social biases in Large Language Models (LLMs) by analyzing their outputs. However, the classifiers used in analysis often have inherent biases, leading to unfair conclusions. This study examines such biases in open-generation benchmarks like BOLD and SAGED. Using the MGSD dataset, we conduct two experiments. The first uses counterfactuals to measure prediction variations across demographic groups by altering stereotype-related prefixes. The second applies explainability tools (SHAP) to validate that the observed biases stem from these counterfactuals. Results reveal unequal treatment of demographic descriptors, calling for more robust bias metric models. |
| title | Assessing Bias in Metric Models for LLM Open-Ended Generation Bias Benchmarks |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2410.11059 |