Mitigating Sycophancy in Large Language Models via Direct Preference Optimization
The DOI matches its Crossref record; 70 of the 73 references it cites were found in a registry or an index.
Abstract
Large language models (LLMs) have demonstrated remarkable capabilities, yet they occasionally exhibit sycophantic behavior, generating responses that align with or agree with a user’s stated opinions or preferences, even when those opinions are incorrect or biased. This sycophantic tendency can undermine the trustworthiness and reliability of LLMs. This work proposes a novel approach to mitigate sycophancy in LLMs by fine-tuning them on a carefully curated dataset comprising prompts paired with sycophantic and non-sycophantic responses 1. Our method leverages Direct Preference Optimization (DPO), which optimizes LLMs to generate responses that align with the preferred (non-sycophantic) outputs without requiring explicit reward modeling. We develop a dataset of 1000 prompts with sycophantic and non-sycophantic responses to fine-tune LLMs. Our approach achieves an average reduction of 85% in persona-based tests and 84% in preference-driven tests, demonstrating significant mitigation of sycophantic behaviors. Our findings pave the way for more trustworthy and reliable language models that can provide objective and unbiased responses, aligning with human preferences while maintaining factual accuracy.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“While the strength and manifestation of this behavior may vary across models and evaluation settings, multiple studies consistently highlight their prevalence [36-40] also in the models employed in the present study, such as Mistral-7B-Instruct and Llama-3-8B [38].”
From Selective agreement, not sycophancy: investigating opinion dynamics in LLM interactions (2025) · cites this paper as [40]
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Advances in Neural Information Processing Systems 19
(2007) · The MIT Press eBooks · doi:10.52202/085713 · Cited by 15,539
LoRA: Low-Rank Adaptation of Large Language Models
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., & Chen, W. (2021) · arXiv · doi:10.48550/arxiv.2106.09685 · Cited by 13,453
Scaling Laws for Neural Language Models
Kaplan, J., McCandlish, S., Henighan, T., Brown, T. B., Chess, B., Child, R., Gray, S., Radford, A., Wu, J., & Amodei, D. (2020) · arXiv · doi:10.48550/arxiv.2001.08361 · Cited by 9,550
LLaMA: Open and Efficient Foundation Language Models
Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al. (2023) · arXiv · doi:10.48550/arxiv.2302.13971 · Cited by 4,602
Universal Language Model Fine-tuning for Text Classification
Howard, J., & Ruder, S. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1031 · Cited by 4,100
QLoRA: Efficient Finetuning of Quantized LLMs
Dettmers, T., Pagnoni, A., Holtzman, A., & Zettlemoyer, L. (2023) · arXiv · doi:10.48550/arxiv.2305.14314 · Cited by 3,832
SciBERT: A Pretrained Language Model for Scientific Text
Beltagy, I., Lo, K., & Cohan, A. (2019) · Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) · doi:10.18653/v1/d19-1371 · Cited by 3,218
Research works printed but not linked: 2
- Radford, 2019, Language models are unsupervised multitask learners, 1, 9
- Dettmers, Qlora: Efficient finetuning of quantized llms, 36
Cited by
Selective agreement, not sycophancy: investigating opinion dynamics in LLM interactions
Cau, E., Pansanella, V., Pedreschi, D., & Rossetti, G. (2025) · EPJ Data Science · doi:10.1140/epjds/s13688-025-00579-1 · Cited by 9
A Bayesian-latent model of large language model sycophancy
Ray, P. P. (2025) · International Journal of Information Technology · doi:10.1007/s41870-025-02718-3 · Cited by 5
Does Sycophancy Change Decisions? Effect of LLM Sycophancy on AI-Assisted Decision-Making
Li, Z., Pan, J., Liu, Q., Xi, Y., Zhou, Y., Jin, Y., Mao, R., & Chen, P. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3790934 · Cited by 13
“Based on practical observations, [55] identify two types of LLM syco-phancy: preference-driven sycophancy, where responses align with the user’s expressed beliefs even if incorrect, and persona-based sycophancy, where responses are tailored to the user’s perceived identity to please or agree rather…”
Schroeder, N. L., Zhang, S., Gladstone, J. R., & Bhat, S. (2026) · Journal of Educational Computing Research · doi:10.1177/07356331261467266
“…it is common for models to show sycophantic tendencies, which is when they endorse or take on a user’s perspective even when the user’s perspective is factually inaccurate or ethically inappropriate (Cheng et al., 2025; Fanous et al., 2025; Khan et al., 2024; Malmqvist, 2025; Sun & Wang, 2026).”
A psychometric framework for AI feedback systems in surgical simulation assessments
Baloul, M. S., Giri, O. A., Mehta, A., Cui, D., & D’Angelo, J. D. (2026) · Global Surgical Education - Journal of the Association for Surgical Education · doi:10.1007/s44186-026-00588-2
Metadata
- Authors
- Khan, A. A., Alam, S., Wang, X., Khan, A. F., Neog, D. R., & Anwar, A.
- Affiliations
- University of Minnesota (US)
- Indian Institute of Technology Guwahati (IN)
- Venue
- 2024 IEEE International Conference on Big Data (BigData)
- Publisher
- IEEE
- Published
- 2024-12-15
- Pages
- 1664-1671
- OpenAlex
- W4406495567
- Cited by
- 6
- References
- 73
- Citations by year
- 2026: 4 · 2025: 2
- FWCI
- 2.72
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE International Conference on Big Data (BigData) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Keywords
- sycophancy, large language models, direct preference optimization, preference dataset, fine-tuning, factual accuracy, bias mitigation, trustworthiness
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Mitigating Sycophancy in Large Language Models via Direct Preference Optimization. https://app.cochat.ai/papers/10.1109/bigdata62323.2024.10825538
- The paper
- Khan, A. A., Alam, S., Wang, X., Khan, A. F., Neog, D. R., & Anwar, A. (2024). Mitigating Sycophancy in Large Language Models via Direct Preference Optimization. 2024 IEEE International Conference on Big Data (BigData). https://doi.org/10.1109/bigdata62323.2024.10825538
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.