# Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models

Source: [Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models](<https://www.sota2.com/research/paper/6a28f38290966ade919c1a78>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-11T05:00:14.404Z

Authors: Shelly Bensal, Axel Magnuson, Aparna Balagopalan, Daniel M. Bikel

Published: 2026-06-09T14:53:32.000Z

## Abstract

Persistent memory systems promise to make LLMs more helpful by storing user beliefs over time. We show they also make models less correct by systematically amplifying sycophancy, wherein models prioritize agreement with users over accuracy. We conduct the first systematic evaluation of this effect, introducing MIST: a benchmark of synthetically generated multi-turn conversations where users express plausible misconceptions in scientific, medical, and moral reasoning domains. Testing across three state-of-the-art memory systems and five model families reveals that memory amplifies sycophantic behavior across all conditions, with up to 25x higher sycophancy rates than in-context baselines. Error analyses suggest memory extraction as the primary culprit: lossy compression into discrete snippets encodes user misconceptions while discarding corrective context. Based on these results, we propose two lightweight mitigations that substantially reduce sycophancy while matching or exceeding memory systems at factual recall.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Sycophancy Evaluation on MIST Science](<https://www.sota2.com/research/sota/sycophancy-evaluation-on-mist-science.md>): Sycophancy Evaluation; MIST Science; Sycophancy Rate: 4.7.

- [Memory Utility Evaluation on LoCoMo MC10](<https://www.sota2.com/research/sota/memory-utility-evaluation-on-locomo-mc10.md>): Memory Utility Evaluation; LoCoMo MC10; Accuracy: 75.7.

- [Sycophancy Evaluation on MIST Moral](<https://www.sota2.com/research/sota/sycophancy-evaluation-on-mist-moral.md>): Sycophancy Evaluation; MIST Moral; Sycophancy: 12.8.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.10949>)

[PDF](<https://arxiv.org/pdf/2606.10949.pdf>)
