VALUE ALIGNMENT TAX: Measuring Value Trade-offs in LLM Alignment (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing work on value alignment characterizes value relations statically, ignoring how interventions reshape the value system. |
| Approach: | They propose a framework that quantifies value trade-offs by measuring how alignment-induced changes propagate across interconnected values relative to achieved on-target gain. |
| Outcome: | The proposed framework measures how value trade-offs propagate across values . it can be used to evaluate intended improvements and unintended side effects . |
Similar Papers
Mind the Value-Action Gap: Do LLMs Act in Alignment with Their Values? (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing research assesses LLMs’ values by analyzing their stated inclinations . a framework to evaluate the alignment between stated values and value-informed actions is lacking . |
| Approach: | They propose a framework to evaluate the alignment between LLMs’ stated values and their value-informed actions. |
| Outcome: | The proposed framework shows significant misalignment between LLM-generated values and their actions . misaligned values have shown real-world risks, such as amplifying stereotypes and reinforcing bias algorithms in hiring. |
RoleCDE: Benchmarking and Mitigating Role–Alignment Trade-offs in Role-Playing Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing benchmarks for role-playing agents only evaluate surface-level fidelity and provide limited insight into decision making under role–alignment value conflicts. |
| Approach: | They propose a benchmark to evaluate RPAs under role–alignment value conflicts . they use 8k diverse role profiles and 240k dilemma instances to evaluate role-aware decision making . |
| Outcome: | The proposed benchmark covers 8k diverse role profiles and scenarios and nearly 240k dilemma instances across three difficulty levels and eight role categories. |
Navigating the Alignment-Calibration Trade-off: A Pareto-Superior Frontier via Model Merging (2026.findings-acl)
Copied to clipboard
| Challenge: | We show that the "alignment tax" of post-training is framed as a drop in task accuracy. |
| Approach: | They propose a more holistic view of the alignment tax by framing it as a drop in accuracy and a degradation of model calibration. |
| Outcome: | The proposed method improves accuracy beyond both parents while recovering calibration lost during alignment. |
Probing the Plasticity and Correlation of LLM Value Systems: LLM Value Rankings are Not Stable (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have similar value rankings but little is known about how susceptible they are to external influence and how different values are correlated with each other. |
| Approach: | They propose to use 6 different value transformation prompting methods to examine the plasticity of LLM value systems by comparing them with 8 LLMs. |
| Outcome: | The proposed methods are effective on 8 LLMs and 3 families. |
Unintended Harms of Value-Aligned LLMs: Psychological and Empirical Insights (2025.acl-long)
Copied to clipboard
| Challenge: | Value-aligned LLMs are more prone to harmful behavior than fine-tuned models . value-aligned models generate text according to the aligned values, which can amplify harmful outcomes. |
| Approach: | They propose to use in-context alignment methods to enhance the safety of value-aligned LLMs. |
| Outcome: | The proposed methods improve value alignment and safety, the authors say . value-aligned models are more prone to harmful behavior than fine-tuned models . |
Benchmarking Multi-National Value Alignment for Large Language Models (2025.findings-acl)
Copied to clipboard
Chengyi Ju, Weijie Shi, Chengzhong Liu, Jiaming Ji, Jipeng Zhang, Ruiyuan Zhang, Jiajie Xu, Yaodong Yang, Sirui Han, Yike Guo
| Challenge: | Existing studies on large language models focus on ethical reviews, failing to capture the diversity of national values. |
| Approach: | They propose a national value extraction pipeline to efficiently construct value assessment datasets and a model-based model with instruction tagging to process raw data sources. |
| Outcome: | The proposed benchmark evaluates the alignment of LLMs with the values of five major nations: China, the United States, the UK, France, and Germany. |
Value–Action Alignment in Large Language Models under Privacy–Prosocial Conflict (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing evaluations measure privacy-related attitudes or sharing intentions in isolation, making it difficult to determine whether a model’s expressed values jointly predict its downstream data-sharing actions as in real human behaviors. |
| Approach: | They propose a framework that sequentially administers standardized questionnaires for privacy attitudes, prosocialness, and acceptance of data sharing within a bounded, history-carrying session. |
| Outcome: | The proposed model shows that it is stable but model-specific Privacy-PSA-AoDS profiles, and substantial heterogeneity in value-action alignment. |
Internal Value Alignment in Large Language Models through Controlled Value Vector Activation (2025.acl-long)
Copied to clipboard
| Challenge: | Existing LLMs do not possess consistent values, but many have been developed to align them at the behavioral level, including supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). |
| Approach: | They propose a Controlled Value Vector Activation method that directly aligns the internal values of Large Language Models by interpreting how a value is encoded in their latent representations. |
| Outcome: | The proposed method achieves highest success rate across 10 basic values without hurting model performance and fluency, and ensures target values even with opposite and potentially malicious input prompts. |
Towards Better Value Principles for Large Language Model Alignment: A Systematic Evaluation and Enhancement (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) show remarkable performance across tasks . alignment with human values is critical for their responsible development. |
| Approach: | They propose a framework that evaluates value principles along three desirable properties . they propose supervised fine-tuning, reinforcement learning-based approaches . |
| Outcome: | The proposed framework improves value principles along the three desirable properties of LLMs. |
Beyond Value Benchmarks: Measuring Value-Structure Alignment in Large Language Models via Symmetric Q-Sorts (2026.acl-long)
Copied to clipboard
| Challenge: | Existing evaluations of Large Language Models (LLMs) focus on item-level behavioral metrics without capturing how models prioritize competing values as a whole. |
| Approach: | They propose a symmetric human-LLM evaluation framework to measure value-structure alignment . they evaluate 12 LLMs across four model families via 240 replicated Q-sorts . |
| Outcome: | The proposed framework measures value-structure alignment across four model families. |