# Mechanistic Origin of Moral Indifference in Language Models

arXiv: https://arxiv.org/abs/2603.15615 · HTML: https://mlsanity.com/paper/2603.15615/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 40 references that are in the map.
- First posted on arXiv 16 Mar 2026.

## Builds on (most cited first, 27 of 27)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings](https://mlsanity.com/paper/1909.00512/) (2019)
- [Aligning AI With Shared Human Values](https://mlsanity.com/paper/2008.02275/) (2020)
- [Persona Vectors: Monitoring and Controlling Character Traits in Language Models](https://mlsanity.com/paper/2507.21509/) (2025)
- [Language Models Represent Space and Time](https://mlsanity.com/paper/2310.02207/) (2023)
- [Social Chemistry 101: Learning to Reason about Social and Moral Norms](https://mlsanity.com/paper/2011.00620/) (2020)
- [Agentic Misalignment: How LLMs Could Be Insider Threats](https://mlsanity.com/paper/2510.05179/) (2025)
- [Benchmarking Complex Instruction-Following with Multiple Constraints Composition](https://mlsanity.com/paper/2407.03978/) (2024)
- [From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery](https://mlsanity.com/paper/2508.14111/) (2025)
- [Values in the Wild: Discovering and Analyzing Values in Real-World Language Model Interactions](https://mlsanity.com/paper/2504.15236/) (2025)
- [MoralBench: Moral Evaluation of LLMs](https://mlsanity.com/paper/2406.04428/) (2024)
- [Fake Alignment: Are LLMs Really Aligned Well?](https://mlsanity.com/paper/2311.05915/) (2023)
- [Flames: Benchmarking Value Alignment of LLMs in Chinese](https://mlsanity.com/paper/2311.06899/) (2023)
- [From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning](https://mlsanity.com/paper/2505.17117/) (2025)
- [OpenRT: An Open-Source Red Teaming Framework for Multimodal LLMs](https://mlsanity.com/paper/2601.01592/) (2026)
- [Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges](https://mlsanity.com/paper/2507.19672/) (2025)
- [Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models](https://mlsanity.com/paper/2511.15304/) (2025)
- [Reflection-Bench: Evaluating Epistemic Agency in Large Language Models](https://mlsanity.com/paper/2410.16270/) (2024)
- [The Other Mind: How Language Models Exhibit Human Temporal Cognition](https://mlsanity.com/paper/2507.15851/) (2025)

## Abstract (arXiv)

Existing behavioral alignment techniques for Large Language Models (LLMs) often neglect the discrepancy between surface compliance and internal unaligned representations, leaving LLMs vulnerable to long-tail risks. More crucially, we posit that LLMs possess an inherent state of moral indifference due to compressing distinct moral concepts into uniform probability distributions. We verify and remedy this indifference in LLMs' latent representations, utilizing 251k moral vectors constructed upon Prototype Theory and the Social-Chemistry-101 dataset. Firstly, our analysis across 23 models reveals that current LLMs fail to represent the distinction between opposed moral categories and fine-grained typicality gradients within these categories; notably, neither model scaling, architecture, nor explicit alignment reshapes this indifference. We then employ Sparse Autoencoders on Qwen3-8B, isolate mono-semantic moral features, and targetedly reconstruct their topological relationships to align with ground-truth moral vectors. This representational alignment naturally improves moral reasoning and granularity, achieving a 75% pairwise win-rate on the independent adversarial Flames benchmark. Finally, we elaborate on the remedial nature of current intervention methods from an experientialist philosophy, arguing that endogenously aligned AI might require a transformation from post-hoc corrections to proactive cultivation.