Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908577177796608 |
|---|---|
| author | Zhang, Yunfan McKeown, Kathleen Muresan, Smaranda |
| author_facet | Zhang, Yunfan McKeown, Kathleen Muresan, Smaranda |
| contents | Large Language Models (LLMs) are typically trained to reflect a relatively uniform set of values, which limits their applicability to tasks that require understanding of nuanced human perspectives. Recent research has underscored the importance of enabling LLMs to support steerable pluralism -- the capacity to adopt a specific perspective and align generated outputs with it. In this work, we investigate whether Chain-of-Thought (CoT) reasoning techniques can be applied to building steerable pluralistic models. We explore several methods, including CoT prompting, fine-tuning on human-authored CoT, fine-tuning on synthetic explanations, and Reinforcement Learning with Verifiable Rewards (RLVR). We evaluate these approaches using the Value Kaleidoscope and OpinionQA datasets. Among the methods studied, RLVR consistently outperforms others and demonstrates strong training sample efficiency. We further analyze the generated CoT traces with respect to faithfulness and safety. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_04045 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment Zhang, Yunfan McKeown, Kathleen Muresan, Smaranda Computation and Language Machine Learning Large Language Models (LLMs) are typically trained to reflect a relatively uniform set of values, which limits their applicability to tasks that require understanding of nuanced human perspectives. Recent research has underscored the importance of enabling LLMs to support steerable pluralism -- the capacity to adopt a specific perspective and align generated outputs with it. In this work, we investigate whether Chain-of-Thought (CoT) reasoning techniques can be applied to building steerable pluralistic models. We explore several methods, including CoT prompting, fine-tuning on human-authored CoT, fine-tuning on synthetic explanations, and Reinforcement Learning with Verifiable Rewards (RLVR). We evaluate these approaches using the Value Kaleidoscope and OpinionQA datasets. Among the methods studied, RLVR consistently outperforms others and demonstrates strong training sample efficiency. We further analyze the generated CoT traces with respect to faithfulness and safety. |
| title | Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2510.04045 |