Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.10415 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914562418147328 |
|---|---|
| author | Lu, Junyu Ji, Deyi Liu, Xuanyi Zhu, Lanyun Xu, Bo Yang, Liang Hua, Xian-Sheng Lin, Hongfei |
| author_facet | Lu, Junyu Ji, Deyi Liu, Xuanyi Zhu, Lanyun Xu, Bo Yang, Liang Hua, Xian-Sheng Lin, Hongfei |
| contents | Large language models for subjectivity analysis are typically trained with aggregated labels, which compress variations in human judgment into a single supervision signal. This paradigm overlooks the intrinsic uncertainty of low-agreement samples and often induces overconfident predictions, undermining reliability and generalization in complex subjective settings. In this work, we advocate uncertainty-aware subjectivity analysis, where models are expected to make predictions while expressing uncertainty that reflects human disagreement. To operationalize this perspective, we propose a two-phase Disagreement Perception and Uncertainty Alignment (DPUA) framework. Specifically, DPUA jointly models label prediction, rationale generation, and uncertainty expression under an uncertainty-aware setting. In the disagreement perception phase, adaptive decoupled learning enhances the model's sensitivity to disagreement-related cues while preserving task performance. In the uncertainty alignment phase, GRPO-based reward optimization further improves uncertainty-aware reasoning and aligns the model's confidence expression with the human disagreement distribution. Experiments on three subjectivity analysis tasks show that DPUA preserves task performance while better aligning model uncertainty with human disagreement, mitigating overconfidence on boundary samples, and improving out-of-distribution generalization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_10415 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Aligning LLM Uncertainty with Human Disagreement in Subjectivity Analysis Lu, Junyu Ji, Deyi Liu, Xuanyi Zhu, Lanyun Xu, Bo Yang, Liang Hua, Xian-Sheng Lin, Hongfei Computation and Language Large language models for subjectivity analysis are typically trained with aggregated labels, which compress variations in human judgment into a single supervision signal. This paradigm overlooks the intrinsic uncertainty of low-agreement samples and often induces overconfident predictions, undermining reliability and generalization in complex subjective settings. In this work, we advocate uncertainty-aware subjectivity analysis, where models are expected to make predictions while expressing uncertainty that reflects human disagreement. To operationalize this perspective, we propose a two-phase Disagreement Perception and Uncertainty Alignment (DPUA) framework. Specifically, DPUA jointly models label prediction, rationale generation, and uncertainty expression under an uncertainty-aware setting. In the disagreement perception phase, adaptive decoupled learning enhances the model's sensitivity to disagreement-related cues while preserving task performance. In the uncertainty alignment phase, GRPO-based reward optimization further improves uncertainty-aware reasoning and aligns the model's confidence expression with the human disagreement distribution. Experiments on three subjectivity analysis tasks show that DPUA preserves task performance while better aligning model uncertainty with human disagreement, mitigating overconfidence on boundary samples, and improving out-of-distribution generalization. |
| title | Aligning LLM Uncertainty with Human Disagreement in Subjectivity Analysis |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2605.10415 |