---
title: "Where meaning lives — paper by Tikhomirova and Wulff (2026)"
description: "This study asks at which depth transformer language models make psychologically meaningful word properties most… 37 references."
authors: ["Taisiia Tikhomirova", "Dirk U. Wulff"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.1008/"
doi: "10.18653/v1/2026.findings-acl.1008"
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/tikhomirova-2026-where-meaning-lives-layer-wise/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/tikhomirova-2026-where-meaning-lives-layer-wise/)

[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/56/) / Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language models

# Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language models

**Authors:** Taisiia Tikhomirova and [Dirk U. Wulff](https://md.research.elamur.ai/free/en/authors/dirk-u-wulff/) · **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.1008/) · [PDF](https://aclanthology.org/2026.findings-acl.1008.pdf) · [DOI 10.18653/v1/2026.findings-acl.1008](https://doi.org/10.18653/v1/2026.findings-acl.1008) · [OpenAlex](https://openalex.org/W7166859656)

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

## Summary
This study asks at which depth transformer language models make psychologically meaningful word properties most accessible, and whether the answer depends on how embeddings are extracted. The authors took 58 word-level features from the psychNorms metabase, which aggregates human ratings, behavioral task data and corpus statistics. They built a shared set of 9,966 English words with high overlap across features. Ten openly available models were probed: five encoders and five large decoders. For each layer, the authors fit linear probes and measured feature-selective performance. Embeddings came from three methods: the word alone, a fixed template, or an average over natural contexts.

Where features appear to be localized depends strongly on the extraction method. Contextualized embeddings give higher selectivity and different layer profiles. Final layers are seldom the best place to read out this information. Even so, models agree on a relative depth order: lexical properties peak early, while experiential and affective ones peak later. Single-method, final-layer analyses can therefore mislead both interpretability claims and practical feature extraction.

## Key points
- **Features and words:** 58 high-coverage features (N>4600 words each) were taken from psychNorms (Hussain et al., 2024).
  - Categories include frequency, valence, arousal, concreteness, age of acquisition, sensory–motor and lexical-decision measures.
  - The 9,966-word set has a median SUBTLEXUS frequency rank of 9,201.
  - It is 52% nouns, 22% verbs and 19% adjectives.
- **Encoder models:** BERT Large, RoBERTa Large, DeBERTa-v3 Large, BGE-M3 and Jina-v3, each with 24 layers.
- **Decoder models:** Mistral-24B, Phi-4, GPT-OSS-20B, Gemma-3-27B and Qwen3-32B, with 24 to 64 layers.
- **Extraction methods:** isolated, template and averaged-context embeddings were compared.
  - For decoders, peak selectivity was about 0.54–0.67 with isolated words.
  - With averaged contexts it rose to about 0.69–0.75.
- **Stated limitations:**
  - The study covers English only.
  - It uses linear probes, so results reflect how accessible information is, not whether it is present.
  - It offers no causal evidence.
  - Many features are correlated with each other.
  - Architecture effects are confounded with differences in post-training.

*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

- [University of Basel](https://md.research.elamur.ai/free/en/universities/university-of-basel/) — University, Basel, Switzerland · 2 works · 71 citations

Also on the byline: [Max Planck Institute for Human Development](https://ror.org/02pp7px91).

📚 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 · [🧠 Neuroscience](https://md.research.elamur.ai/free/en/shelves/neuroscience/) · 186 works · [🧠 Cognitive Neuroscience](https://md.research.elamur.ai/free/en/shelves/neuroscience/cognitive-neuroscience/) · 185 works

🧭 Topics (2): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works · [🧒 Psycholinguistics & language acquisition](https://md.research.elamur.ai/free/en/topics/psycholinguistics/) · 108 works

## Cite

Cite as: Taisiia Tikhomirova and Dirk U. Wulff. 2026. Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language models. 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{tikhomirova2026where,
  title = {{Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language models}},
  author = {Taisiia Tikhomirova and Dirk U. Wulff},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1008},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1008}
}
```

- [← previous Where do LLMs currently stand on biomedical NER in both clean and noisy settings ?](https://md.research.elamur.ai/free/en/papers/ye-2026-where-do-llms-currently-stand/)
- [next → Where Paths Split: Localized, Calibrated Control of Moral Reasoning in Large Language Models](https://md.research.elamur.ai/free/en/papers/yuan-2026-where-paths-split-localized-calibrated/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/56/)

## References

37 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Marah Abdin, Jyoti Aneja, Harkirat Behl, Sébastien Bubeck, Ronen Eldan, Suriya Gunasekar, Michael Harrison, Russell J. Hewett, Mojan Javaheripi, Piero Kauffmann, James R. Lee, Yin Tat Lee, Yuanzhi Li, Weishung Liu, Caio C. T. Mendes, Anh Nguyen, Eric Price, Gustavo de Rosa, Olli Saarikivi, and 8 oth… — [Phi-4 Technical Report](https://md.research.elamur.ai/free/en/papers/abdin-2024-phi-4-technical-report/)
2. Yonatan Belinkov. 2022. [Probing classifiers: Promises, shortcomings, and advances](https://md.research.elamur.ai/free/en/papers/belinkov-2022-probing-classifiers-promises-shortcomings-advances/). Computational Linguistics, 48(1):207–219.
3. Rishi Bommasani, Kelly Davis, and Claire Cardie. 2020. [Interpreting Pretrained Contextualized Representations via Reductions to Static Embeddings](https://md.research.elamur.ai/free/en/papers/bommasani-2020-interpreting-pretrained-contextualized-representations-reductions/). In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4758– 4781, Online. Association for Computational Lingu…
4. Ting-Yun Chang and Yun-Nung Chen. 2019. What does this word mean? explaining contextualized embeddings with natural language definition. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing… [find](https://scholar.google.com/scholar?q=Ting%2DYun+Chang+and+Yun%2DNung+Chen%2E+2019%2E+What+does+this+word+mean%3F+explaining+contextualized+embeddings+with+natural+language+definition%2E+In+Proceedings+of+the+2019+Conference+on+Empirical+Methods+in+N)
5. Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. [Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation](https://md.research.elamur.ai/free/en/papers/chen-2024-m3-embedding-multi-linguality/). Preprint, arXiv:2402.03216.
6. Gabriella Chronis and Katrin Erk. 2020. When is a bishop not like a rook? when it’s like a rabbi! multi-prototype BERT embeddings for estimating semantic relationships. In Proceedings of the 24th Conference on Computational Natural Language Learning, pages 227–244, Online. Association for Computatio… [find](https://scholar.google.com/scholar?q=Gabriella+Chronis+and+Katrin+Erk%2E+2020%2E+When+is+a+bishop+not+like+a+rook%3F+when+it%E2%80%99s+like+a+rabbi%21+multi%2Dprototype+BERT+embeddings+for+estimating+semantic+relationships%2E+In+Proceedings+of+the+24th+Conf)
7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [BERT: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,…
8. Kawin Ethayarajh. 2019. How contextual are contextualized word representations? Comparing the geometry of BERT, ELMo, and GPT-2 embeddings. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Process… [find](https://scholar.google.com/scholar?q=Kawin+Ethayarajh%2E+2019%2E+How+contextual+are+contextualized+word+representations%3F+Comparing+the+geometry+of+BERT%2C+ELMo%2C+and+GPT%2D2+embeddings%2E+In+Proceedings+of+the+2019+Conference+on+Empirical+Methods+i)
9. Wes Gurnee and Max Tegmark. 2024. models represent space and time. arXiv:2310.02207. [find](https://scholar.google.com/scholar?q=Wes+Gurnee+and+Max+Tegmark%2E+2024%2E+models+represent+space+and+time%2E+arXiv%3A2310%2E02207%2E)
10. Language Preprint, Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2023. [Debertav3: Improving deberta using electra-style pre-training with gradient-disentangled embedding sharing](https://md.research.elamur.ai/free/en/papers/he-2021-debertav3-improving-deberta-using-electra/). Preprint, arXiv:2111.09543.
11. John Hewitt and Percy Liang. 2019. Designing and interpreting probes with control tasks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2733–2743, Hong Kong, Chi… [find](https://scholar.google.com/scholar?q=John+Hewitt+and+Percy+Liang%2E+2019%2E+Designing+and+interpreting+probes+with+control+tasks%2E+In+Proceedings+of+the+2019+Conference+on+Empirical+Methods+in+Natural+Language+Processing+and+the+9th+Internati)
12. John Hewitt and Christopher D. Manning. 2019. [A structural probe for finding syntax in word representations](https://md.research.elamur.ai/free/en/papers/hewitt-2019-structural-probe-finding-syntax-word/). In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4129…
13. Zak Hussain, Rui Mata, Ben R. Newell, and Dirk U. Wulff. 2024. Probing the contents of semantic representations from text, behavior, and brain data using the psychnorms metabase. arXiv preprint arXiv:2412.04936. — [arXiv:2412.04936](https://arxiv.org/abs/2412.04936) [find](https://scholar.google.com/scholar?q=Zak+Hussain%2C+Rui+Mata%2C+Ben+R%2E+Newell%2C+and+Dirk+U%2E+Wulff%2E+2024%2E+Probing+the+contents+of+semantic+representations+from+text%2C+behavior%2C+and+brain+data+using+the+psychnorms+metabase%2E+arXiv+preprint+arXiv%3A)
14. Ganesh Jawahar, Benoît Sagot, and Djamé Seddah. 2019. [What does BERT learn about the structure of language?](https://md.research.elamur.ai/free/en/papers/jawahar-2019-what-does-bert-learn-about/) In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3651–3657, Florence, Italy. Association for Computational Linguistics.
15. Bastien Liétard, Mostafa Abdou, and Anders Søgaard. 2021. Do language models know the way to Rome? In Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP, pages 510–517, Punta Cana, Dominican Republic. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Bastien+Li%C3%A9tard%2C+Mostafa+Abdou%2C+and+Anders+S%C3%B8gaard%2E+2021%2E+Do+language+models+know+the+way+to+Rome%3F+In+Proceedings+of+the+Fourth+BlackboxNLP+Workshop+on+Analyzing+and+Interpreting+Neural+Networks+for+N)
16. Yanai Elazar, Shauli Ravfogel, Alon Jacovi, and Yoav Goldberg. 2021. [Amnesic probing: Behavioral explanation with amnesic counterfactuals](https://md.research.elamur.ai/free/en/papers/elazar-2021-amnesic-probing-behavioral-explanation-amnesic/). Transactions of the Association for Computational Linguistics, 9:160– 175.
17. Yongjie Lin, Yi Chern Tan, and Robert Frank. 2019. [Open sesame: Getting inside BERT’s linguistic knowledge](https://md.research.elamur.ai/free/en/papers/lin-2019-open-sesame-getting-inside-bert/). In Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 241–253, Florence, Italy. Association for Computational Linguistics.
18. Jeffrey L Elman. 2004. An alternative view of the mental lexicon. Trends in cognitive sciences, 8(7):301– 306. [find](https://scholar.google.com/scholar?q=Jeffrey+L+Elman%2E+2004%2E+An+alternative+view+of+the+mental+lexicon%2E+Trends+in+cognitive+sciences%2C+8%287%29%3A301%E2%80%93+306%2E)
19. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. [find](https://scholar.google.com/scholar?q=Yinhan+Liu%2C+Myle+Ott%2C+Naman+Goyal%2C+Jingfei+Du%2C+Mandar+Joshi%2C+Danqi+Chen%2C+Omer+Levy%2C+Mike+Lewis%2C+Luke+Zettlemoyer%2C+and+Veselin+Stoyanov%2E+2019%2E)
20. [Roberta: A robustly optimized bert pretraining approach](https://md.research.elamur.ai/free/en/papers/liu-2019-roberta-robustly-optimized-bert-pretraining/). Preprint, arXiv:1907.11692.
21. Zhu Liu, Cunliang Kong, Ying Liu, and Maosong Sun. 2024. Fantastic semantics and where to find them: Investigating which layers of generative llms reflect lexical semantics. arXiv preprint arXiv:2403.01509. [find](https://scholar.google.com/scholar?q=Zhu+Liu%2C+Cunliang+Kong%2C+Ying+Liu%2C+and+Maosong+Sun%2E+2024%2E+Fantastic+semantics+and+where+to+find+them%3A+Investigating+which+layers+of+generative+llms+reflect+lexical+semantics%2E+arXiv+preprint+arXiv%3A2403%2E)
22. Mistral AI Team. 2025. Mistral small 3. [https://mistral.ai/news/mistral-small-3](https://mistral.ai/news/mistral-small-3). Accessed: 2025-12-05. [↗](https://mistral.ai/news/mistral-small-3)
23. OpenAI, :, Sandhini Agarwal, Lama Ahmad, Jason Ai, Sam Altman, Andy Applebaum, Edwin Arbus, Rahul K. Arora, Yu Bai, Bowen Baker, Haiming Bao, Boaz Barak, Ally Bennett, Tyler Bertao, Nivedita Brett, Eugene Brevdo, Greg Brockman, Sebastien Bubeck, and 108 others. 2025. [gpt-oss-120b and gpt-oss-20b mod…](https://md.research.elamur.ai/free/en/papers/agarwal-2025-gpt-oss-120b-gpt-oss/)
24. Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller. 2019. [Language models as knowledge bases?](https://md.research.elamur.ai/free/en/papers/petroni-2019-language-models-as-knowledge-bases/) In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural…
25. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. [Exploring the limits of transfer learning with a unified text-to-text transformer](https://md.research.elamur.ai/free/en/papers/raffel-2019-exploring-limits-transfer-learning-unified/). Journal of Machine Learning Research, 21(140):1–67.
26. Abhilasha Ravichander, Yonatan Belinkov, and Eduard Hovy. 2021. [Probing the probing paradigm: Does probing accuracy entail task relevance?](https://md.research.elamur.ai/free/en/papers/ravichander-2021-probing-probing-paradigm-does-probing/) In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 3363–3377, Online. Association…
27. Oscar Skean, Md Rifat Arefin, Dan Zhao, Niket Patel, Jalal Naghiyev, Yann LeCun, and Ravid Shwartz-Ziv. 2025. [Layer by layer: Uncovering hidden representations in language models](https://md.research.elamur.ai/free/en/papers/skean-2025-layer-layer-uncovering-hidden-representations/). Preprint, arXiv:2502.02013.
28. Saba Sturua, Isabelle Mohr, Mohammad Kalim Akram, Michael Günther, Bo Wang, Markus Krimmel, Feng Wang, Georgios Mastrapas, Andreas Koukounas, Nan Wang, and Han Xiao. 2024. [jina-embeddings-v3: Multilingual embeddings with task lora](https://md.research.elamur.ai/free/en/papers/sturua-2024-jina-embeddings-v3-multilingual-embeddings/). Preprint, arXiv:2409.10173.
29. Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, Louis Rouillard, Thomas Mesnard, Geoffrey Cideron, Jean bastien Grill, Sabela Ramos, Edouard Yvinec, Michelle Casbon, Etienne Pot, Ivo Penche… [find](https://scholar.google.com/scholar?q=Gemma+Team%2C+Aishwarya+Kamath%2C+Johan+Ferret%2C+Shreya+Pathak%2C+Nino+Vieillard%2C+Ramona+Merhej%2C+Sarah+Perrin%2C+Tatiana+Matejovicova%2C+Alexandre+Ram%C3%A9%2C+Morgane+Rivi%C3%A8re%2C+Louis+Rouillard%2C+Thomas+Mesnard%2C+Geoffrey)
30. Ian Tenney, Dipanjan Das, and Ellie Pavlick. 2019. [BERT rediscovers the classical NLP pipeline](https://md.research.elamur.ai/free/en/papers/tenney-2019-bert-rediscovers-classical-nlp-pipeline/). In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4593– 4601, Florence, Italy. Association for Computational Linguistics.
31. Andreas Waldis, Yotam Perlitz, Leshem Choshen, Yufang Hou, and Iryna Gurevych. 2024. Holmes: A benchmark to assess the linguistic competence of language models. Transactions of the Association for Computational Linguistics, 12:1616–1647. [find](https://scholar.google.com/scholar?q=Andreas+Waldis%2C+Yotam+Perlitz%2C+Leshem+Choshen%2C+Yufang+Hou%2C+and+Iryna+Gurevych%2E+2024%2E+Holmes%3A+A+benchmark+to+assess+the+linguistic+competence+of+language+models%2E+Transactions+of+the+Association+for+Com)
32. Dirk U Wulff, Simon De Deyne, Michael N Jones, and Rui Mata. 2019. New perspectives on the aging lexicon. Trends in cognitive sciences, 23(8):686–698. [find](https://scholar.google.com/scholar?q=Dirk+U+Wulff%2C+Simon+De+Deyne%2C+Michael+N+Jones%2C+and+Rui+Mata%2E+2019%2E+New+perspectives+on+the+aging+lexicon%2E+Trends+in+cognitive+sciences%2C+23%288%29%3A686%E2%80%93698%2E)
33. Ningyu Xu, Qi Zhang, Chao Du, Qiang Luo, Xipeng Qiu, Xuanjing Huang, and Menghan Zhang. 2025. Revealing emergent human-like conceptual representations from language prediction. Proceedings of the National Academy of Sciences, 122(44):e2512514122. [find](https://scholar.google.com/scholar?q=Ningyu+Xu%2C+Qi+Zhang%2C+Chao+Du%2C+Qiang+Luo%2C+Xipeng+Qiu%2C+Xuanjing+Huang%2C+and+Menghan+Zhang%2E+2025%2E+Revealing+emergent+human%2Dlike+conceptual+representations+from+language+prediction%2E+Proceedings+of+the+Nati)
34. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. [Qwen3 technical report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). Preprint, arXiv:2505.09388.
35. Fred Zhang and Neel Nanda. 2024. [Towards best practices of activation patching in language models: Metrics and methods](https://md.research.elamur.ai/free/en/papers/zhang-2023-towards-best-practices-activation-patching/). Preprint, arXiv:2309.16042.
36. Jingxiang Zhang and Lujia Zhong. 2025. Decoding emotion in the deep: A systematic study of how llms represent, retain, and express emotion. Preprint, arXiv:2510.04064. — [arXiv:2510.04064](https://arxiv.org/abs/2510.04064) [find](https://scholar.google.com/scholar?q=Jingxiang+Zhang+and+Lujia+Zhong%2E+2025%2E+Decoding+emotion+in+the+deep%3A+A+systematic+study+of+how+llms+represent%2C+retain%2C+and+express+emotion%2E+Preprint%2C+arXiv%3A2510%2E04064%2E+%E2%80%94+%5BarXiv%3A2510%2E04064%5D%28https%3A%2F%2Farx)
37. Wentao Zhu, Zhining Zhang, and Yizhou Wang. 2024. Language models represent beliefs of self and others. Preprint, arXiv:2402.18496. — [arXiv:2402.18496](https://arxiv.org/abs/2402.18496) [find](https://scholar.google.com/scholar?q=Wentao+Zhu%2C+Zhining+Zhang%2C+and+Yizhou+Wang%2E+2024%2E+Language+models+represent+beliefs+of+self+and+others%2E+Preprint%2C+arXiv%3A2402%2E18496%2E+%E2%80%94+%5BarXiv%3A2402%2E18496%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2402%2E18496%29)

*Updated 2 November 2026*