Benchmarking Overton Pluralism in LLMs
This work addresses the need for scalable evaluation tools for pluralistic alignment in LLMs, which is crucial for model development but incremental in establishing a measurable benchmark.
The paper tackles the problem of measuring diverse viewpoint representation in LLMs by introducing OvertonScore, a set coverage metric, and finds that models achieve scores of 0.35-0.41, far below the maximum of 1.0, with DeepSeek V3 performing best.
We introduce a novel framework for measuring Overton pluralism in LLMs--the extent to which diverse viewpoints are represented in model outputs. We (i) formalize Overton pluralism as a set coverage metric (OvertonScore), (ii) conduct a large-scale U.S.-representative human study (N = 1209; 60 questions; 8 LLMs), and (iii) develop an automated benchmark that closely reproduces human judgments. On average, models achieve OvertonScores of 0.35--0.41, with DeepSeek V3 performing best; yet all models remain far below the theoretical maximum of 1.0, revealing substantial headroom for improvement. Because repeated large-scale human studies are costly and slow, scalable evaluation tools are essential for model development. Hence, we propose an automated benchmark that achieves high rank correlation with human judgments ($ρ=0.88$), providing a practical proxy without replacing human assessment. By turning pluralistic alignment from a normative aim into a measurable benchmark, our work establishes a foundation for systematic progress toward more pluralistic LLMs.