Paper
Think Before You Lie: How Reasoning Improves Honesty
Authors
Ann Yuan, Asma Ghandeharioun, Carter Blum, Alicia Machado, Jessica Hoffmann, Daphne Ippolito, Martin Wattenberg, Lucas Dixon, Katja Filippova
Abstract
While existing evaluations of large language models (LLMs) measure deception rates, the underlying conditions that give rise to deceptive behavior are poorly understood. We investigate this question using a novel dataset of realistic moral trade-offs where honesty incurs variable costs. Contrary to humans, who tend to become less honest given time to deliberate (Capraro, 2017; Capraro et al., 2019), we find that reasoning consistently increases honesty across scales and for several LLM families. This effect is not only a function of the reasoning content, as reasoning traces are often poor predictors of final behaviors. Rather, we show that the underlying geometry of the representational space itself contributes to the effect. Namely, we observe that deceptive regions within this space are metastable: deceptive answers are more easily destabilized by input paraphrasing, output resampling, and activation noise than honest ones. We interpret the effect of reasoning in this vein: generating deliberative tokens as part of moral reasoning entails the traversal of a biased representational space, ultimately nudging the model toward its more stable, honest defaults.
Metadata
Related papers
Gen-Searcher: Reinforcing Agentic Search for Image Generation
Kaituo Feng, Manyuan Zhang, Shuang Chen, Yunlong Lin, Kaixuan Fan, Yilei Jian... • 2026-03-30
On-the-fly Repulsion in the Contextual Space for Rich Diversity in Diffusion Transformers
Omer Dahary, Benaya Koren, Daniel Garibi, Daniel Cohen-Or • 2026-03-30
Graphilosophy: Graph-Based Digital Humanities Computing with The Four Books
Minh-Thu Do, Quynh-Chau Le-Tran, Duc-Duy Nguyen-Mai, Thien-Trang Nguyen, Khan... • 2026-03-30
ParaSpeechCLAP: A Dual-Encoder Speech-Text Model for Rich Stylistic Language-Audio Pretraining
Anuj Diwan, Eunsol Choi, David Harwath • 2026-03-30
RAD-AI: Rethinking Architecture Documentation for AI-Augmented Ecosystems
Oliver Aleksander Larsen, Mahyar T. Moghaddam • 2026-03-30
Raw Data (Debug)
{
"raw_xml": "<entry>\n <id>http://arxiv.org/abs/2603.09957v1</id>\n <title>Think Before You Lie: How Reasoning Improves Honesty</title>\n <updated>2026-03-10T17:52:49Z</updated>\n <link href='https://arxiv.org/abs/2603.09957v1' rel='alternate' type='text/html'/>\n <link href='https://arxiv.org/pdf/2603.09957v1' rel='related' title='pdf' type='application/pdf'/>\n <summary>While existing evaluations of large language models (LLMs) measure deception rates, the underlying conditions that give rise to deceptive behavior are poorly understood. We investigate this question using a novel dataset of realistic moral trade-offs where honesty incurs variable costs. Contrary to humans, who tend to become less honest given time to deliberate (Capraro, 2017; Capraro et al., 2019), we find that reasoning consistently increases honesty across scales and for several LLM families. This effect is not only a function of the reasoning content, as reasoning traces are often poor predictors of final behaviors. Rather, we show that the underlying geometry of the representational space itself contributes to the effect. Namely, we observe that deceptive regions within this space are metastable: deceptive answers are more easily destabilized by input paraphrasing, output resampling, and activation noise than honest ones. We interpret the effect of reasoning in this vein: generating deliberative tokens as part of moral reasoning entails the traversal of a biased representational space, ultimately nudging the model toward its more stable, honest defaults.</summary>\n <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n <published>2026-03-10T17:52:49Z</published>\n <arxiv:primary_category term='cs.AI'/>\n <author>\n <name>Ann Yuan</name>\n </author>\n <author>\n <name>Asma Ghandeharioun</name>\n </author>\n <author>\n <name>Carter Blum</name>\n </author>\n <author>\n <name>Alicia Machado</name>\n </author>\n <author>\n <name>Jessica Hoffmann</name>\n </author>\n <author>\n <name>Daphne Ippolito</name>\n </author>\n <author>\n <name>Martin Wattenberg</name>\n </author>\n <author>\n <name>Lucas Dixon</name>\n </author>\n <author>\n <name>Katja Filippova</name>\n </author>\n </entry>"
}