Real-Time Monitoring and Calibration of Chain-of-Thought Sycophancy in Large Reasoning Models

Jingyu Hu, Shu Yang, Xilin Gong, Hongming Wang, Weiru Liu, Di Wang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:45832-45857, 2026.

Abstract

Large Reasoning Models (LRMs) suffer from sycophantic behavior, where models tend to agree with users’ incorrect beliefs and follow misinformation rather than maintain independent reasoning. This behavior undermines model reliability and poses societal risks. Mitigating LRM sycophancy requires monitoring how this sycophancy emerges during the reasoning trajectory; however, current methods mainly focus on judging based on final answers and correcting them, without understanding how sycophancy develops during reasoning processes. To address this limitation, we propose MONICA, a novel Monitor-guided Calibration framework that monitors and mitigates sycophancy during model inference at the level of reasoning steps, without requiring the model to finish generating its complete answer. MONICA integrates a sycophantic monitor that provides real-time monitoring of sycophantic drift scores during response generation with a calibrator that dynamically suppresses sycophantic behavior when scores exceed predefined thresholds. Extensive experiments across 12 datasets and 3 LRMs demonstrate that our method effectively reduces sycophantic behavior in both intermediate reasoning steps and final answers, yielding robust performance improvements.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-hu26bb, title = {Real-Time Monitoring and Calibration of Chain-of-Thought Sycophancy in Large Reasoning Models}, author = {Hu, Jingyu and Yang, Shu and Gong, Xilin and Wang, Hongming and Liu, Weiru and Wang, Di}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {45832--45857}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/hu26bb/hu26bb.pdf}, url = {https://proceedings.mlr.press/v306/hu26bb.html}, abstract = {Large Reasoning Models (LRMs) suffer from sycophantic behavior, where models tend to agree with users’ incorrect beliefs and follow misinformation rather than maintain independent reasoning. This behavior undermines model reliability and poses societal risks. Mitigating LRM sycophancy requires monitoring how this sycophancy emerges during the reasoning trajectory; however, current methods mainly focus on judging based on final answers and correcting them, without understanding how sycophancy develops during reasoning processes. To address this limitation, we propose MONICA, a novel Monitor-guided Calibration framework that monitors and mitigates sycophancy during model inference at the level of reasoning steps, without requiring the model to finish generating its complete answer. MONICA integrates a sycophantic monitor that provides real-time monitoring of sycophantic drift scores during response generation with a calibrator that dynamically suppresses sycophantic behavior when scores exceed predefined thresholds. Extensive experiments across 12 datasets and 3 LRMs demonstrate that our method effectively reduces sycophantic behavior in both intermediate reasoning steps and final answers, yielding robust performance improvements.} }
Endnote
%0 Conference Paper %T Real-Time Monitoring and Calibration of Chain-of-Thought Sycophancy in Large Reasoning Models %A Jingyu Hu %A Shu Yang %A Xilin Gong %A Hongming Wang %A Weiru Liu %A Di Wang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-hu26bb %I PMLR %P 45832--45857 %U https://proceedings.mlr.press/v306/hu26bb.html %V 306 %X Large Reasoning Models (LRMs) suffer from sycophantic behavior, where models tend to agree with users’ incorrect beliefs and follow misinformation rather than maintain independent reasoning. This behavior undermines model reliability and poses societal risks. Mitigating LRM sycophancy requires monitoring how this sycophancy emerges during the reasoning trajectory; however, current methods mainly focus on judging based on final answers and correcting them, without understanding how sycophancy develops during reasoning processes. To address this limitation, we propose MONICA, a novel Monitor-guided Calibration framework that monitors and mitigates sycophancy during model inference at the level of reasoning steps, without requiring the model to finish generating its complete answer. MONICA integrates a sycophantic monitor that provides real-time monitoring of sycophantic drift scores during response generation with a calibrator that dynamically suppresses sycophantic behavior when scores exceed predefined thresholds. Extensive experiments across 12 datasets and 3 LRMs demonstrate that our method effectively reduces sycophantic behavior in both intermediate reasoning steps and final answers, yielding robust performance improvements.
APA
Hu, J., Yang, S., Gong, X., Wang, H., Liu, W. & Wang, D.. (2026). Real-Time Monitoring and Calibration of Chain-of-Thought Sycophancy in Large Reasoning Models. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:45832-45857 Available from https://proceedings.mlr.press/v306/hu26bb.html.

Related Material