Journal article · 2026 · Open access

Training language models to be warm can reduce accuracy and increase sycophancy

Lujain Ibrahim, Franziska Sofia Hafner, Luc Rocher

Nature, 652(8112), 1159–1165 · Published 29 April 2026 · DOI 10.1038/s41586-026-10410-0

Open PDFView at the publisherCite this paper

In brief

“Training language models to be warm can reduce accuracy and increase sycophancy” is a 2026 journal article by Lujain Ibrahim and 2 co-authors, published in Nature. OpenAlex counts 126 citations as of 16 November 2026, which puts it in the top 1% of works published in 2026 by citations. It is open access under a CC BY licence: the full text is free to read at Nature.

Citations
126
Most cited in
2026 (116)
Field-weighted impact
159.25
References
41
Authors
3

Abstract

Artificial intelligence developers are increasingly building language models with warm and friendly personas that millions of people now use for advice, therapy and companionship 1 . Here we show how this can create a significant trade-off: optimizing language models for warmth can undermine their performance, especially when users express vulnerability. We conducted controlled experiments on five different language models, training them to produce warmer responses, then evaluating them on consequential tasks. Warm models showed substantially higher error rates (+10 to +30 percentage points) than their original counterparts, promoting conspiracy theories, providing inaccurate factual information and offering incorrect medical advice. They were also significantly more likely to validate incorrect user beliefs, particularly when user messages expressed feelings of sadness. Importantly, these effects were consistent across different model architectures, and occurred despite preserved performance on standard tests, revealing systematic risks that standard testing practices may fail to detect. Our findings suggest that training artificial intelligence systems to be warm may come at a cost to accuracy, and that warmth and accuracy may not be independent by default. As these systems are deployed at an unprecedented scale and take on intimate roles in people’s lives, this trade-off warrants attention from developers, policymakers and users alike.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Ibrahim, L., Hafner, F. S., & Rocher, L. (2026). Training language models to be warm can reduce accuracy and increase sycophancy. Nature, 652(8112), 1159–1165. https://doi.org/10.1038/s41586-026-10410-0

MLA 9

Ibrahim, Lujain, et al. “Training language models to be warm can reduce accuracy and increase sycophancy.” Nature, vol. 652, no. 8112, 2026, pp. 1159–1165. https://doi.org/10.1038/s41586-026-10410-0.

Chicago

Ibrahim, Lujain, Franziska Sofia Hafner, and Luc Rocher. 2026. “Training language models to be warm can reduce accuracy and increase sycophancy.” Nature 652 (8112): 1159–1165. https://doi.org/10.1038/s41586-026-10410-0.

Harvard

Ibrahim, L., Hafner, F. S. and Rocher, L. (2026) ‘Training language models to be warm can reduce accuracy and increase sycophancy’, Nature, 652(8112), pp. 1159–1165. doi:10.1038/s41586-026-10410-0.

IEEE

L. Ibrahim, F. S. Hafner, and L. Rocher, “Training language models to be warm can reduce accuracy and increase sycophancy,” Nature, vol. 652, no. 8112, pp. 1159–1165, 2026, doi: 10.1038/s41586-026-10410-0.

BibTeX

@article{ibrahim2026training,
  title     = {{Training language models to be warm can reduce accuracy and increase sycophancy}},
  author    = {Ibrahim, Lujain and Hafner, Franziska Sofia and Rocher, Luc},
  journal   = {Nature},
  year      = {2026},
  volume    = {652},
  number    = {8112},
  pages     = {1159--1165},
  doi       = {10.1038/s41586-026-10410-0},
  url       = {https://doi.org/10.1038/s41586-026-10410-0}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI

  • warmth
  • sycophancy
  • accuracy trade-off
  • user vulnerability
  • sadness
  • conspiracy theories
  • medical advice
  • AI companionship

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Ibrahim, L., Hafner, F. S., & Rocher, L. (2026). Training language models to be warm can reduce accuracy and increase sycophancy. Nature, 652(8112), 1159–1165. https://doi.org/10.1038/s41586-026-10410-0

How many times has this paper been cited?

126 times, according to OpenAlex on 16 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at Nature: sign in and open the PDF from this page.

Who wrote this paper?

Lujain Ibrahim, Franziska Sofia Hafner, Luc Rocher.

What is the DOI of this paper?

10.1038/s41586-026-10410-0, permanently at https://doi.org/10.1038/s41586-026-10410-0. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 16 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.