Value Entanglement: Conflation Between Different Kinds of Good In (Some) Large Language Models
On the map: LLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Abstract (arXiv)
Value alignment of Large Language Models (LLMs) requires us to empirically measure these models' actual, acquired representation of value. Among the characteristics of value representation in humans is that they distinguish among value of different kinds. We investigate whether LLMs likewise distinguish three different kinds of good: moral, grammatical, and economic. By probing model behavior, embeddings, and residual stream activations, we report pervasive cases of value entanglement: a conflation between these distinct representations of value. Specifically, both grammatical and economic valuation was found to be overly influenced by moral value, relative to human norms. This conflation was repaired by selective ablation of the activation vectors associated with morality.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 11 of its 21 references that are in the map. First posted on arXiv 22 Feb 2026.
Builds on
11 of 21 references.
References · 11
- 2025Qwen3 Technical Reportcited by 8,195
- 2024Qwen2.5 Technical Reportcited by 4,713
- 2025Gemma 3 Technical Reportcited by 1,686
- 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
- 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
- 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,009
- 2024Refusal in Language Models Is Mediated by a Single Directioncited by 960
- 2025Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMscited by 158
- 2025Convergent Linear Representations of Emergent Misalignmentcited by 66
- 2025Utility Engineering: Analyzing and Controlling Emergent Value Systems in AIscited by 63
- 2025Shared Parameter Subspaces and Cross-Task Linearity in Emergently Misaligned Behaviorcited by 7
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Do LLMs Have Values? A Quantitative Analysis and Alignment Framework for Values in Large Language Models
- 2026Reward Models Inherit Value Biases from Pretraining
- 2023Heterogeneous Value Alignment Evaluation for Large Language Models
- 2026Beyond Value Benchmarks: Measuring Value-Structure Alignment in Large Language Models via Symmetric Q-Sorts
- 2025Do Language Models Think Consistently? A Study of Value Preferences Across Varying Response Lengths
- 2024Exploring Multilingual Concepts of Human Value in Large Language Models: Is Value Alignment Consistent, Transferable and Controllable across Languages?
- 2025Cultural Value Alignment in Large Language Models: A Prompt-based Analysis of Schwartz Values in Gemini, ChatGPT, and DeepSeek
- 2025Differences in the Moral Foundations of Large Language Models