# Analysing Self-Harm Representations in Language Models: a Cross-Architecture Study

arXiv: https://arxiv.org/abs/2607.21988 · HTML: https://mlsanity.com/paper/2607.21988/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 36 references that are in the map.
- First posted on arXiv 24 Jul 2026.

## Builds on (most cited first, 19 of 19)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [Probing Classifiers: Promises, Shortcomings, and Advances](https://mlsanity.com/paper/2102.12452/) (2021)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [Programming Refusal with Conditional Activation Steering](https://mlsanity.com/paper/2409.05907/) (2024)
- [Depression and Self-Harm Risk Assessment in Online Forums](https://mlsanity.com/paper/1709.01848/) (2017)
- [Safety is Not Only About Refusal: Reasoning-Enhanced Fine-tuning for Interpretable LLM Safety](https://mlsanity.com/paper/2503.05021/) (2025)
- [Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs](https://mlsanity.com/paper/2503.05371/) (2025)
- [Towards Preemptive Detection of Depression and Anxiety in Twitter](https://mlsanity.com/paper/2011.05249/) (2020)
- [Identifying Depression on Twitter](https://mlsanity.com/paper/1607.07384/) (2016)
- [Just a Scratch: Enhancing LLM Capabilities for Self-harm Detection through Intent Differentiation and Emoji Interpretation](https://mlsanity.com/paper/2506.05073/) (2025)
- [Sensitive Content Classification in Social Media: A Holistic Resource and Evaluation](https://mlsanity.com/paper/2411.19832/) (2024)
- [Domain-Specific Constitutional AI: Enhancing Safety in LLM-Powered Mental Health Chatbots](https://mlsanity.com/paper/2509.16444/) (2025)
- [Preventing Another Tessa: Modular Safety Middleware For Health-Adjacent AI Assistants](https://mlsanity.com/paper/2509.07022/) (2025)

## Abstract (arXiv)

Self-harm content is particularly challenging to detect using NLP techniques, and is also a high-stakes task which requires the highest accuracy to enable timely intervention or flagging at-risk users. We therefore present an analysis of how LLMs represent such self-harm content, which has downstream applications in self-harm detection, LLM intervention and governance and policing. In this paper, we focus on two datasets and four models, and perform two main experiments: (1) We train and evaluate linear probes across all layers of each model on two self-harm datasets: X-Sensitive and SH-Detection. Across both corpora, self-harm information crystallizes in the final 3 - 7% of network layers (93 to 97% depth). (2) We extract contrastive self-harm directions and, after performing a normaliation step, we find that the most accurate probes are not necessarily the most linearly separable. In particular, we find Gemma-3-4B to represent this textitcontrastive self-harm direction in a slightly different, more intricate way than the other LLMs.