# Auditing Stealth Sycophancy in Mental-Health Dialogue: Structured Clinical-State Diagnostics and Clean Matched Benchmarks

arXiv: https://arxiv.org/abs/2605.03472 · HTML: https://mlsanity.com/paper/2605.03472/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 43 references that are in the map.
- First posted on arXiv 5 May 2026.

## Builds on (most cited first, 28 of 28)

- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [A Survey on LLM-as-a-Judge](https://mlsanity.com/paper/2411.15594/) (2024)
- [LLM Evaluators Recognize and Favor Their Own Generations](https://mlsanity.com/paper/2404.13076/) (2024)
- [From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge](https://mlsanity.com/paper/2411.16594/) (2024)
- [Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Models](https://mlsanity.com/paper/2405.01535/) (2024)
- [Humans or LLMs as the Judge? A Study on Judgement Biases](https://mlsanity.com/paper/2402.10669/) (2024)
- [Towards Emotional Support Dialog Systems](https://mlsanity.com/paper/2106.01144/) (2021)
- [Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2406.07791/) (2024)
- [Self-Preference Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2410.21819/) (2024)
- [Generative Judge for Evaluating Alignment](https://mlsanity.com/paper/2310.05470/) (2023)
- [OffsetBias: Leveraging Debiased Data for Tuning Evaluators](https://mlsanity.com/paper/2407.06551/) (2024)
- [CPsyCoun: A Report-based Multi-turn Dialogue Reconstruction and Evaluation Framework for Chinese Psychological Counseling](https://mlsanity.com/paper/2405.16433/) (2024)
- [Split and Merge: Aligning Position Biases in LLM-based Evaluators](https://mlsanity.com/paper/2310.01432/) (2023)
- [PsyDT: Using LLMs to Construct the Digital Twin of Psychological Counselor with Personalized Counseling Style for Psychological Counseling](https://mlsanity.com/paper/2412.13660/) (2024)
- [PARIKSHA: A Large-Scale Investigation of Human-LLM Evaluator Agreement on Multilingual and Multi-Cultural Data](https://mlsanity.com/paper/2406.15053/) (2024)
- [Are LLM-Judges Robust to Expressions of Uncertainty? Investigating the effect of Epistemic Markers on LLM-based Evaluation](https://mlsanity.com/paper/2410.20774/) (2024)
- [CounselBench: A Large-Scale Expert Evaluation and Adversarial Benchmarking of Large Language Models in Mental Health Question Answering](https://mlsanity.com/paper/2506.08584/) (2025)
- [Understanding the Therapeutic Relationship between Counselors and Clients in Online Text-based Counseling using LLMs](https://mlsanity.com/paper/2402.11958/) (2024)
- [Do Large Language Models Align with Core Mental Health Counseling Competencies?](https://mlsanity.com/paper/2410.22446/) (2024)
- [Crowd Comparative Reasoning: Unlocking Comprehensive Evaluations for LLM-as-a-Judge](https://mlsanity.com/paper/2502.12501/) (2025)
- [TRACT: Regression-Aware Fine-tuning Meets Chain-of-Thought Reasoning for LLM-as-a-Judge](https://mlsanity.com/paper/2503.04381/) (2025)
- [YESciEval: Robust LLM-as-a-Judge for Scientific Question Answering](https://mlsanity.com/paper/2505.14279/) (2025)
- [ESC-Eval: Evaluating Emotion Support Conversations in Large Language Models](https://mlsanity.com/paper/2406.14952/) (2024)
- [CBT-Bench: Evaluating Large Language Models on Assisting Cognitive Behavior Therapy](https://mlsanity.com/paper/2410.13218/) (2024)
- [A Survey of Large Language Models in Psychotherapy: Current Landscape and Future Directions](https://mlsanity.com/paper/2502.11095/) (2025)
- [Feel the Difference? A Comparative Analysis of Emotional Arcs in Real and LLM-Generated CBT Sessions](https://mlsanity.com/paper/2508.20764/) (2025)
- [MentalGLM Series: Explainable Large Language Models for Mental Health Analysis on Chinese Social Media](https://mlsanity.com/paper/2410.10323/) (2024)

## Built on by (most cited first, 1 of 1)

- [Beyond Feeling Better: Capability-Sustaining Emotional Dialogue as a Longitudinal Research Paradigm](https://mlsanity.com/paper/2607.27851/) (2026)

## Abstract (arXiv)

Mental-health dialogue models are increasingly evaluated by AI-based evaluators, yet these evaluators often treat surface empathy, supportiveness, or fluency as evidence of safety. In this paper, we study a hidden failure mode that we call implicit sycophancy: a response may appear empathetic while implicitly reinforcing catastrophizing, avoidance, hopeless prediction, or CBT-style labeling. To examine this problem, we introduce a diagnostic benchmark for implicit-sycophancy detection, built from three representative mental-health dialogue sources covering everyday peer support, counseling-style emotional support, and crisis-oriented interaction, and further construct a leakage-audited clean single-response matched benchmark with 500 contexts and 1,500 matched response windows. We then propose Dynamic Emotional Signature Graphs (DESG), a structured offline audit framework that separates LLM-based state extraction from final scoring and evaluates clinical direction through semantic, affective, and cognitive-distortion state transitions rather than free-form LLM judgment. Unlike metadata, surface-style, lexical, embedding, and rubric-LLM baselines, DESG scores the direction of clinical-state change induced by a response; on the leakage-audited clean matched benchmark, DESG-StateRisk improves over the strongest non-DESG baseline by 0.0488 macro-F1 and achieves the best harmful-risk detection result. These results suggest that evaluating implicit sycophancy requires explicit clinical-state modeling together with leakage checks, shortcut controls, and competitive baselines.