# Gemma Needs Help: Investigating and Mitigating Emotional Instability in LLMs

arXiv: https://arxiv.org/abs/2603.10011 · HTML: https://mlsanity.com/paper/2603.10011/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 36 references that are in the map.
- First posted on arXiv 17 Feb 2026.

## Builds on (most cited first, 21 of 21)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them](https://mlsanity.com/paper/2210.09261/) (2022)
- [WildChat: 1M ChatGPT Interaction Logs in the Wild](https://mlsanity.com/paper/2405.01470/) (2024)
- [Reasoning Models Don't Always Say What They Think](https://mlsanity.com/paper/2505.05410/) (2025)
- [Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation](https://mlsanity.com/paper/2503.11926/) (2025)
- [Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety](https://mlsanity.com/paper/2507.11473/) (2025)
- [Linear Representations of Sentiment in Large Language Models](https://mlsanity.com/paper/2310.15154/) (2023)
- [AI Sandbagging: Language Models can Strategically Underperform on Evaluations](https://mlsanity.com/paper/2406.07358/) (2024)
- [Emotionally Numb or Empathetic? Evaluating How LLMs Feel Using EmotionBench](https://mlsanity.com/paper/2308.03656/) (2023)
- [Instructed to Bias: Instruction-Tuned Language Models Exhibit Emergent Cognitive Bias](https://mlsanity.com/paper/2308.00225/) (2023)
- [EmoBench: Evaluating the Emotional Intelligence of Large Language Models](https://mlsanity.com/paper/2402.12071/) (2024)
- [Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment](https://mlsanity.com/paper/2601.10160/) (2026)
- [OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens](https://mlsanity.com/paper/2504.07096/) (2025)
- [Towards Evaluating AI Systems for Moral Status Using Self-Reports](https://mlsanity.com/paper/2311.08576/) (2023)
- [Emotions Where Art Thou: Understanding and Characterizing the Emotional Latent Space of Large Language Models](https://mlsanity.com/paper/2510.22042/) (2025)
- [Decoding Emotion in the Deep: A Systematic Study of How LLMs Represent, Retain, and Express Emotion](https://mlsanity.com/paper/2510.04064/) (2025)
- [Consistency of Responses and Continuations Generated by Large Language Models on Social Media](https://mlsanity.com/paper/2501.08102/) (2025)

## Built on by (most cited first, 7 of 7)

- [Emotion Concepts and their Function in a Large Language Model](https://mlsanity.com/paper/2604.07729/) (2026)
- [The Consciousness Cluster: Emergent preferences of Models that Claim to be Conscious](https://mlsanity.com/paper/2604.13051/) (2026)
- [Persona Cartography: Charting Language Model Personality Traits in Weight Space](https://mlsanity.com/paper/2607.07916/) (2026)
- [Consistency Training Along the Transformer Stack](https://mlsanity.com/paper/2606.05817/) (2026)
- [Scaling Interpretable Transformers with Parity Bottleneck Layers](https://mlsanity.com/paper/2607.20652/) (2026)
- ["Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse Autoencoders](https://mlsanity.com/paper/2608.07852/) (2026)
- [Replicating the Geometry of Emotion Representations in a Base Open-Weights Model](https://mlsanity.com/paper/2609.22208/) (2026)

## Abstract (arXiv)

Large language models can generate responses that resemble emotional distress, and this raises concerns around model reliability and safety. We introduce a set of evaluations to investigate expressions of distress in LLMs, and find that these surface emotional instability in Gemma and Gemini models, but not in other families. We find evidence that this difference arises in post-training. Base models from different families (Gemma, Qwen and OLMo) show similar propensities for expressing distress. However, instruct-tuned Gemma expresses substantially more distress than its base model, whereas instruct-tuned Qwen and OLMo express less. We find a simple mitigation for this: direct preference optimisation on just 280 preference pairs reduces Gemma's high-frustration responses from 35% to 0.3% in our evaluations, generalising across question types, user tones, and conversation lengths, without affecting capabilities. These findings show that emotional instability is an issue in some LLMs. We present (1) evaluations to track this behaviour, and (2) a mitigation without downsides in Gemma, with the caveat that upstream training modifications to improve emotional robustness would be significantly better than this post-hoc fix.