---
title: "Choose Your Lens — paper by Zhao et al. (2026)"
description: "The authors propose VGC-CoT (Value-Graph-Consistent Chain-of-Thought), a neuro-symbolic framework that separates facts… 32 references."
authors: ["Gejian Zhao", "Hanzhou Wu", "Xinpeng Zhang"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.2075/"
doi: "10.18653/v1/2026.findings-acl.2075"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/zhao-2026-choose-your-lens-multi-perspective/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhao-2026-choose-your-lens-multi-perspective/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/26/) / Choose Your Lens: Multi-Perspective Value Alignment of Chain-of-Thought Reasoning

# Choose Your Lens: Multi-Perspective Value Alignment of Chain-of-Thought Reasoning

**Authors:** Gejian Zhao, Hanzhou Wu and Xinpeng Zhang · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.2075/) · [PDF](https://aclanthology.org/2026.findings-acl.2075.pdf) · [DOI 10.18653/v1/2026.findings-acl.2075](https://doi.org/10.18653/v1/2026.findings-acl.2075) · [OpenAlex](https://openalex.org/W7166820378)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
LLMs asked to argue from a specific viewpoint often fall into motivated reasoning. Instead of weighing values differently, they invent supporting "facts," such as claiming rent control expands housing supply. The authors propose VGC-CoT (Value-Graph-Consistent Chain-of-Thought), a neuro-symbolic framework that separates facts from values. Factual claims should stay identical across perspectives, while value priorities may differ. The framework uses a composite graph with three parts:
- a shared, human-verified causal layer built from Wikipedia;
- per-perspective value layers grounded in the Schwartz value ontology;
- cross-layer links from factual outcomes to value concepts.

For each question and perspective, a canonical subgraph is retrieved. The model's generated CoT is then mapped onto the graph through information extraction. Factual claims without graph support count as hallucinations. Graph-based alignment scores derived from this mapping drive both preference training and reranking of candidate CoTs at inference time. On the GRAPH-POLICYQA benchmark with LLaMA-3-8B, the method cuts factual hallucinations roughly threefold versus DPO and steerable baselines. It also raises cross-perspective consistency by about 25%. This suggests models can represent diverse worldviews without bending shared reality.

## Key points
- The VGC-CoT graph has three components: a locked factual layer (G_fact), perspective-specific value layers, and cross-layer mappings. Example perspectives include utilitarian, libertarian and egalitarian.
- The CoT-to-graph projection uses an IE pipeline that extracts triples and grounds them semantically. Value alignment is scored with soft Jaccard similarity against the canonical value subgraph.
- Reported results: about 3× fewer factual hallucinations and +25% cross-perspective consistency compared with DPO and steerable baselines.
- The main experiments use LLaMA-3-8B on GRAPH-POLICYQA. The authors release their datasets but list scaling to larger models (e.g., 70B) as future work.
- Stated limitations:
  - the method depends on graph quality, since graphs were built semi-automatically and verified by graduate students;
  - causal edges are treated as fixed, which is a strong assumption;
  - the factual graph and the value ontology may carry Western-centric bias.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Shanghai University](https://md.research.elamur.ai/free/en/universities/shanghai-university/) — University, Shanghai, China · 6 works · 124 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,954 works · [🤝 Safety Research](https://md.research.elamur.ai/free/en/shelves/social-sciences/safety-research/) · 626 works

🧭 Topics (1): [🧮 Reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/) · 1,685 works

## Cite

Cite as: Gejian Zhao, Hanzhou Wu and Xinpeng Zhang. 2026. Choose Your Lens: Multi-Perspective Value Alignment of Chain-of-Thought Reasoning. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{zhao2026choose,
  title = {{Choose Your Lens: Multi-Perspective Value Alignment of Chain-of-Thought Reasoning}},
  author = {Gejian Zhao, Hanzhou Wu and Xinpeng Zhang},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.2075},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.2075}
}
```

- [← previous ChineseDevBench: A Chinese Developmental Benchmark for Language Development](https://md.research.elamur.ai/free/en/papers/wang-2026-chinesedevbench-chinese-developmental-benchmark-language/)
- [next → ChronoLearn: A GRAG LLM-Based System for Structuring and Exploring Historical Narratives](https://md.research.elamur.ai/free/en/papers/aladdasi-2026-chronolearn-grag-llm-based-system/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/26/)

## References

32 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, and Cameron McKinnon. 2022. [Constitutional ai: Harmlessness from ai feedback](https://md.research.elamur.ai/free/en/papers/bai-2022-constitutional-ai-harmlessness-ai-feedback/). arXiv preprint arXiv:2212.08073.
2. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, and 1 others. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 17682–17690. [find](https://scholar.google.com/scholar?q=Maciej+Besta%2C+Nils+Blach%2C+Ales+Kubicek%2C+Robert+Gerstenberger%2C+and+1+others%2E+2024%2E+Graph+of+thoughts%3A+Solving+elaborate+problems+with+large+language+models%2E+In+Proceedings+of+the+AAAI+Conference+on+Art)
3. Collin Burns, Haotian Ye, Dan Klein, and Jacob Steinhardt. 2023. Discovering latent knowledge in language models without supervision. In Proceedings of the International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Collin+Burns%2C+Haotian+Ye%2C+Dan+Klein%2C+and+Jacob+Steinhardt%2E+2023%2E+Discovering+latent+knowledge+in+language+models+without+supervision%2E+In+Proceedings+of+the+International+Conference+on+Learning+Represe)
4. Yanda Chen, Joe Benton, Ansh Radhakrishnan, Jonathan Uesato, Carson Denison, John Schulman, Arushi Somani, Peter Hase, Misha Wagner, and Fabien Roger. 2025. [Reasoning models don’t always say what they think](https://md.research.elamur.ai/free/en/papers/chen-2025-reasoning-models-don-t-always/). arXiv preprint arXiv:2505.05410.
5. Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. 2024. [Improving factuality and reasoning in language models through multiagent debate](https://md.research.elamur.ai/free/en/papers/du-2023-improving-factuality-reasoning-language-models/). In Forty-first international conference on machine learning.
6. Esin Durmus, Karina Nguyen, Thomas I Liao, Nicholas Schiefer, Amanda Askell, Anton Bakhtin, Carol Chen, Zac Hatfield-Dodds, Danny Hernandez, Nicholas Joseph, and 1 others. 2023. [Towards measuring the representation of subjective global opinions in language models](https://md.research.elamur.ai/free/en/papers/durmus-2023-towards-measuring-representation-subjective-global/). arXiv preprint arXiv:2306.16388.
7. Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, Dasha Metropolitansky, Robert Osazuwa Ness, and Jonathan Larson. 2024. From local to global: A graph rag approach to query-focused summarization. arXiv preprint arXiv:2404.16130. [find](https://scholar.google.com/scholar?q=Darren+Edge%2C+Ha+Trinh%2C+Newman+Cheng%2C+Joshua+Bradley%2C+Alex+Chao%2C+Apurva+Mody%2C+Steven+Truitt%2C+Dasha+Metropolitansky%2C+Robert+Osazuwa+Ness%2C+and+Jonathan+Larson%2E+2024%2E+From+local+to+global%3A+A+graph+rag+app)
8. Ioana Hulpuş, Jonathan Kobbe, Heiner Stuckenschmidt, and Graeme Hirst. 2020. Knowledge graphs meet moral values. In Proceedings of the Joint Conference on Lexical and Computational Semantics, pages 71– 80. [find](https://scholar.google.com/scholar?q=Ioana+Hulpus%CC%A7%2C+Jonathan+Kobbe%2C+Heiner+Stuckenschmidt%2C+and+Graeme+Hirst%2E+2020%2E+Knowledge+graphs+meet+moral+values%2E+In+Proceedings+of+the+Joint+Conference+on+Lexical+and+Computational+Semantics%2C+pages+7)
9. Zhouyu Jiang, Ling Zhong, Mengshu Sun, Jun Xu, Rui Sun, Hui Cai, Shuhan Luo, and Zhiqiang Zhang. 2024. Efficient knowledge infusion via kg–llm alignment. In Findings of the Association for Computational Linguistics, pages 2986–2999. [find](https://scholar.google.com/scholar?q=Zhouyu+Jiang%2C+Ling+Zhong%2C+Mengshu+Sun%2C+Jun+Xu%2C+Rui+Sun%2C+Hui+Cai%2C+Shuhan+Luo%2C+and+Zhiqiang+Zhang%2E+2024%2E+Efficient+knowledge+infusion+via+kg%E2%80%93llm+alignment%2E+In+Findings+of+the+Association+for+Computation)
10. Junfeng Jiao, Saleh Afroogh, Abhejay Murali, Kevin Chen, David Atkinson, and Amit Dhurandhar. 2025a. Llm ethics benchmark: a three-dimensional assessment system for evaluating moral reasoning in large language models. Scientific Reports, 15(1):34642. [find](https://scholar.google.com/scholar?q=Junfeng+Jiao%2C+Saleh+Afroogh%2C+Abhejay+Murali%2C+Kevin+Chen%2C+David+Atkinson%2C+and+Amit+Dhurandhar%2E+2025a%2E+Llm+ethics+benchmark%3A+a+three%2Ddimensional+assessment+system+for+evaluating+moral+reasoning+in+large)
11. Junfeng Jiao, Saleh Afroogh, Yiming Xu, and Connor Phillips. 2025b. Navigating llm ethics: Advancements, challenges, and future directions. AI and Ethics, pages 1–25. [find](https://scholar.google.com/scholar?q=Junfeng+Jiao%2C+Saleh+Afroogh%2C+Yiming+Xu%2C+and+Connor+Phillips%2E+2025b%2E+Navigating+llm+ethics%3A+Advancements%2C+challenges%2C+and+future+directions%2E+AI+and+Ethics%2C+pages+1%E2%80%9325%2E)
12. Shathika Kularatne, Matt Selway, Wolfgang Mayer, and Markus Stumptner. 2025. Automating perdurant meta-property assignment using gpt-4. In Proceedings of the Australasian Joint Conference on Artificial Intelligence, pages 40–53. [find](https://scholar.google.com/scholar?q=Shathika+Kularatne%2C+Matt+Selway%2C+Wolfgang+Mayer%2C+and+Markus+Stumptner%2E+2025%2E+Automating+perdurant+meta%2Dproperty+assignment+using+gpt%2D4%2E+In+Proceedings+of+the+Australasian+Joint+Conference+on+Artificia)
13. Jia Li, Ge Li, Yongmin Li, and Zhi Jin. 2025. Structured chain-of-thought prompting for code generation. ACM Transactions on Software Engineering and Methodology, 34(2):1–23. [find](https://scholar.google.com/scholar?q=Jia+Li%2C+Ge+Li%2C+Yongmin+Li%2C+and+Zhi+Jin%2E+2025%2E+Structured+chain%2Dof%2Dthought+prompting+for+code+generation%2E+ACM+Transactions+on+Software+Engineering+and+Methodology%2C+34%282%29%3A1%E2%80%9323%2E)
14. Hunter Lightman, Vineet Kosaraju, Yuri Burda, Harrison Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. 2023. [Let’s verify step by step](https://md.research.elamur.ai/free/en/papers/lightman-2023-let-s-verify-step-step/). In The twelfth international conference on learning representations.
15. Qing Lyu, Shreya Havaldar, Adam Stein, Li Zhang, Delip Rao, Eric Wong, Marianna Apidianaki, and Chris Callison-Burch. 2023. [Faithful chain-of-thought reasoning](https://md.research.elamur.ai/free/en/papers/lyu-2023-faithful-chain-thought-reasoning/). In Proceedings of IJCNLP-AACL.
16. Taylor Sorensen, Jared Moore, Jillian Fisher, Mitchell Gordon, Niloofar Mireshghallah, Christopher Michael Rytting, Andre Ye, Liwei Jiang, Ximing Lu, and Nouha Dziri. 2024. Position: [A roadmap to pluralistic alignment](https://md.research.elamur.ai/free/en/papers/sorensen-2024-roadmap-pluralistic-alignment/). In Proceedings of the International Conference on Machine Learning, pages 46280–4…
17. Mahmoud Srewa, Tianyu Zhao, and Salma Elmalaki. 2025. Pluralllm: Pluralistic alignment in llms via federated learning. In Proceedings of the International Workshop on Human-Centered Sensing, Modeling, and Intelligent Systems, pages 64–69. [find](https://scholar.google.com/scholar?q=Mahmoud+Srewa%2C+Tianyu+Zhao%2C+and+Salma+Elmalaki%2E+2025%2E+Pluralllm%3A+Pluralistic+alignment+in+llms+via+federated+learning%2E+In+Proceedings+of+the+International+Workshop+on+Human%2DCentered+Sensing%2C+Modeling%2C)
18. Miles Turpin, Julian Michael, Ethan Perez, and Samuel Bowman. 2023. [Language models don’t always say what they think: Unfaithful explanations in chain-of-thought prompting](https://md.research.elamur.ai/free/en/papers/turpin-2023-language-models-don-t-always/). In Advances in Neural Information Processing Systems, volume 36, pages 74952– 74965.
19. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, and 1 others. 2023. Self-refine: Iterative refinement with self-feedback. Advances in neural information processing systems, 36:46534–46594. [find](https://scholar.google.com/scholar?q=Aman+Madaan%2C+Niket+Tandon%2C+Prakhar+Gupta%2C+Skyler+Hallinan%2C+Luyu+Gao%2C+Sarah+Wiegreffe%2C+Uri+Alon%2C+Nouha+Dziri%2C+Shrimai+Prabhumoye%2C+Yiming+Yang%2C+and+1+others%2E+2023%2E+Self%2Drefine%3A+Iterative+refinement+with)
20. Shijie Wang, Wenqi Fan, Yue Feng, Shanru Lin, Xinyu Ma, Shuaiqiang Wang, and Dawei Yin. 2025. [Knowledge graph retrieval-augmented generation for llm-based recommendation](https://md.research.elamur.ai/free/en/papers/wang-2025-knowledge-graph-retrieval-augmented-generation/). In Proceedings of the Annual Meeting of the Association for Computational Linguistics, pages 27152–27168.
21. Lars Malmqvist. 2025. [Sycophancy in large language models: Causes and mitigations](https://md.research.elamur.ai/free/en/papers/malmqvist-2024-sycophancy-large-language-models-causes/). In Proceedings of the Computing Conference, pages 61–74.
22. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V. Le, and Denny Zhou. 2022. [Chain-of-thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). In Advances in Neural Information Processing Systems, volume 35, pages 24824–24837.
23. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, and Alex Ray. 2022. [Training language models to follow instructions with human feedback](https://md.research.elamur.ai/free/en/papers/ouyang-2022-training-language-models-follow-instructions/). Advances in Neural Information Processing Systems, 35:27730–27744.
24. Kentaro Ozeki, Risako Ando, Takanobu Morishita, Hirohiko Abe, Koji Mineshima, and Mitsuhiro Okada. 2025. Normative reasoning in large language models: A comparative benchmark from logical and modal perspectives. In Proceedings of the BlackboxNLP Workshop, pages 276–294. [find](https://scholar.google.com/scholar?q=Kentaro+Ozeki%2C+Risako+Ando%2C+Takanobu+Morishita%2C+Hirohiko+Abe%2C+Koji+Mineshima%2C+and+Mitsuhiro+Okada%2E+2025%2E+Normative+reasoning+in+large+language+models%3A+A+comparative+benchmark+from+logical+and+modal+pe)
25. Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu. 2024. [Unifying large language models and knowledge graphs: A roadmap](https://md.research.elamur.ai/free/en/papers/pan-2023-unifying-large-language-models-knowledge/). IEEE Transactions on Knowledge and Data Engineering, 36(7):3580–3599.
26. Sriyash Poddar, Yanming Wan, Hamish Ivison, Abhishek Gupta, and Natasha Jaques. 2024. Personalizing reinforcement learning from human feedback with variational preference learning. Advances in Neural Information Processing Systems, 37:52516– 52544. [find](https://scholar.google.com/scholar?q=Sriyash+Poddar%2C+Yanming+Wan%2C+Hamish+Ivison%2C+Abhishek+Gupta%2C+and+Natasha+Jaques%2E+2024%2E+Personalizing+reinforcement+learning+from+human+feedback+with+variational+preference+learning%2E+Advances+in+Neural+)
27. Melissa Kazemi Rad, Huy Nghiem, Andy Luo, Sahil Wadhwa, Mohammad Sorower, and Stephen Rawls. 2025. [Refining input guardrails: Enhancing llm-as-a-judge efficiency through chain-of-thought fine-tuning and alignment](https://md.research.elamur.ai/free/en/papers/rad-2025-refining-input-guardrails-enhancing-llm/). arXiv preprint arXiv:2501.13080.
28. Rui Yang, Xiaoman Pan, Feng Luo, Shuang Qiu, Han Zhong, Dong Yu, and Jianshu Chen. 2024. Rewards-in-context: Multi-objective alignment of foundation models with dynamic preference adjustment. In Proceedings of the International Conference on Machine Learning, pages 56276–56297. [find](https://scholar.google.com/scholar?q=Rui+Yang%2C+Xiaoman+Pan%2C+Feng+Luo%2C+Shuang+Qiu%2C+Han+Zhong%2C+Dong+Yu%2C+and+Jianshu+Chen%2E+2024%2E+Rewards%2Din%2Dcontext%3A+Multi%2Dobjective+alignment+of+foundation+models+with+dynamic+preference+adjustment%2E+In+Proce)
29. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. In Advances in Neural Information Processing Systems, volume 36, pages 11809–11822. [find](https://scholar.google.com/scholar?q=Shunyu+Yao%2C+Dian+Yu%2C+Jeffrey+Zhao%2C+Izhak+Shafran%2C+Tom+Griffiths%2C+Yuan+Cao%2C+and+Karthik+Narasimhan%2E+2023%2E+Tree+of+thoughts%3A+Deliberate+problem+solving+with+large+language+models%2E+In+Advances+in+Neural+)
30. Beichen Zhang, Kun Zhou, Xilin Wei, Xin Zhao, Jing Sha, Shijin Wang, and Ji-Rong Wen. 2023. Evaluating and improving tool-augmented computation-intensive math reasoning. Advances in Neural Information Processing Systems, 36:23570–23589. [find](https://scholar.google.com/scholar?q=Beichen+Zhang%2C+Kun+Zhou%2C+Xilin+Wei%2C+Xin+Zhao%2C+Jing+Sha%2C+Shijin+Wang%2C+and+Ji%2DRong+Wen%2E+2023%2E+Evaluating+and+improving+tool%2Daugmented+computation%2Dintensive+math+reasoning%2E+Advances+in+Neural+Information)
31. Yunfan Zhang, Kathleen McKeown, and Smaranda Muresan. 2025. Exploring chain-of-thought reasoning for steerable pluralistic alignment. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 25647–25660. [find](https://scholar.google.com/scholar?q=Yunfan+Zhang%2C+Kathleen+McKeown%2C+and+Smaranda+Muresan%2E+2025%2E+Exploring+chain%2Dof%2Dthought+reasoning+for+steerable+pluralistic+alignment%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natu)
32. Gejian Zhao, Hanzhou Wu, Xinpeng Zhang, and Athanasios V Vasilakos. 2025. Shadowcot: Cognitive hijacking for stealthy reasoning backdoors in llms. arXiv preprint arXiv:2504.05605. — [arXiv:2504.05605](https://arxiv.org/abs/2504.05605) [find](https://scholar.google.com/scholar?q=Gejian+Zhao%2C+Hanzhou+Wu%2C+Xinpeng+Zhang%2C+and+Athanasios+V+Vasilakos%2E+2025%2E+Shadowcot%3A+Cognitive+hijacking+for+stealthy+reasoning+backdoors+in+llms%2E+arXiv+preprint+arXiv%3A2504%2E05605%2E+%E2%80%94+%5BarXiv%3A2504%2E05605%5D)

*Updated 2 November 2026*