Elamur Research⌕
☰
home / papers / 2026 / GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs

GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs

Paper · 2026 · Weidong Tang, Jierui Li, Yueling Hou et al.

Summary

Multimodal LLMs have become better at inferring what a single agent believes or wants. Real social settings demand more: forecasting how private beliefs and goals interact under hierarchy, conformity and information asymmetry, often producing group results nobody intended, such as groupthink or the Abilene Paradox. GroupToM-Bench is presented as the first multimodal benchmark for this group-level Theory of Mind (ToM). Its expert-written scenarios combine visual cues, such as facial expressions or how people are positioned, with dialogue and hidden private states.

Evaluation follows a causal chain with three stages:

  • micro-level BDI, i.e., what each individual believes, desires and intends,
  • meso-level frictions among members and the structural limits the group operates under,
  • macro-level outcome forecasting and attribution of causes.

This chain is operationalized as a seven-level cognitive audit. Frontier models score clearly below humans. They recover individual motives fairly well, yet tend to predict an optimistic, rational consensus, as if a group were simply the sum of its members. The authors call this linear superposition bias and argue that scaling has not removed it. The benchmark therefore serves as a diagnostic tool for socially grounded AI.

Key points

  • The resource contains 240 expert-curated scenarios in eight socio-psychological domains (e.g., information, belief, dilemmas, emotion, conformity, power, identity, inertia), with 3,120 tasks in total.
  • The audit has seven levels. L1–L3 test individual belief, desire and intention before any interaction. L4–L7 cover tension, constraint, collective prediction and mechanistic attribution.
  • Open-ended answers at L5 and L7 are graded by GPT-5 against expert gold references. The authors acknowledge a possible self-scoring bias, since GPT-5 is also among the evaluated models.
  • Visual dependence is uneven across the benchmark. In an image-removal ablation, GPT-4o drops only about 1.9 points on average.
  • Stated limitations include mostly Western social norms in the scenarios and a possible link between safety alignment and models' reluctance to predict collective failure.

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

🎓National University of SingaporeSingapore · Singapore207 works · 39,576 citations🎓University of Science and Technology of ChinaHefei · China145 works · 114,409 citations🎓University of Electronic Science and Technology of ChinaChengdu · China38 works · 2,088 citations🎓Xidian UniversityXi'an · China17 works · 1,357 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

Cite

Cite as Weidong Tang, Jierui Li, Yueling Hou, Zihan Mei, Can Zhang, Xinyan Wan, Zhiyuan Liang, Pengfei Zhou, Yang You and Wangbo Zhao. 2026. GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1859

@inproceedings{tang2026grouptom,
  title = {{GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs}},
  author = {Weidong Tang, Jierui Li, Yueling Hou, Zihan Mei, Can Zhang, Xinyan Wan, Zhiyuan Liang, Pengfei Zhou, Yang You and Wangbo Zhao},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1859},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1859}
}

References

54 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Icek Ajzen. 1991. The theory of planned behavior. Organizational behavior and human decision processes, 50(2):179–211. find
  2. Anthropic. 2025. Introducing claude haiku 4.5. Accessed: 2025-12-20. find
  3. Solomon E Asch. 1956. Studies of independence and conformity: I. a minority of one against a unanimous majority. Psychological monographs: General and applied. find
  4. Simon Baron-Cohen, Alan M Leslie, and Uta Frith. 1985. Does the autistic child have a “theory of mind”? Cognition, 21. find
  5. Matteo Bortoletto, Constantin Ruhdorfer, and Andreas Bulling. 2025. Tom-ssi: Evaluating theory of mind in situated social interactions. In EMNLP, pages 32264–32289. find
  6. Chunkit Chan, Cheng Jiayang, Yauwai Yim, Zheye Deng, Wei Fan, Haoran Li, Xin Liu, Hongming Zhang, Weiqi Wang, and Yangqiu Song. 2024. Nego-tiationtom: A benchmark for stress-testing machine theory of mind on negotiation surrounding. arXiv preprint arXiv:2404.13627. — [arXiv:2404.13627](https://arxiv.org/abs/2404.13627) find
  7. Ruirui Chen, Weifeng Jiang, Chengwei Qin, and Cheston Tan. 2025. Theory of mind in large language models: Assessment and enhancement. In ACL, pages 31539–31558.
  8. Zhuang Chen, Jincenzi Wu, Jinfeng Zhou, Bosi Wen, Guanqun Bi, Gongyao Jiang, Yaru Cao, Mengting Hu, Yunghwei Lai, Zexuan Xiong, and Minlie Huang. 2024. Tombench: Benchmarking theory of mind in large language models. In ACL, pages 15959–15983.
  9. Google DeepMind. 2025. Gemini 3. Accessed: 2025-12-20. find
  10. Ailin Deng, Tri Cao, Zhirui Chen, and Bryan Hooi. 2025. Words or vision: Do vision-language models have blind faith in text? In CVPR, pages 3867–3876. find
  11. Jingtao Ding, Yunke Zhang, Yu Shang, Yuheng Zhang, Zefang Zong, Jie Feng, Yuan Yuan, Hongyuan Su, Nian Li, Nicholas Sukiennik, Fengli Xu, and Yong Li. 2026. Understanding world or predicting future? a comprehensive survey of world models. ACM Comput. Surv., 58:57:1–57:38. find
  12. Xianzhe Fan, Xuhui Zhou, Chuanyang Jin, Kolby Nottingham, Hao Zhu, and Maarten Sap. 2025. Somi-tom: Evaluating multi-perspective theory of mind in embodied social interactions. arXiv preprint arXiv:2506.23046. — [arXiv:2506.23046](https://arxiv.org/abs/2506.23046) find
  13. Mark Granovetter. 1978. Threshold models of collective behavior. American journal of sociology, 83(6):1420– 1443. find
  14. Yuling Gu, Oyvind Tafjord, Hyunwoo Kim, Jared Moore, Ronan Le Bras, Peter Clark, and Yejin Choi. 2024. Simpletom: Exposing the gap between explicit tom inference and implicit tom application in llms. arXiv preprint arXiv:2410.13648. — [arXiv:2410.13648](https://arxiv.org/abs/2410.13648) find
  15. David Ha and Jürgen Schmidhuber. 2018. World models. arXiv preprint arXiv:1803.10122. find
  16. Jerry B Harvey. 1974. The abilene paradox: The management of agreement. Organizational dynamics, 3. find
  17. Irving L Janis. 1972. Victims of groupthink: A psychological study of foreign-policy decisions and fiascoes. Houghton Mifflin. find
  18. Chuanyang Jin, Yutong Wu, Jing Cao, Jiannan Xiang, Yen-Ling Kuo, Zhiting Hu, Tomer Ullman, Antonio Torralba, Joshua Tenenbaum, and Tianmin Shu. 2024. Mmtom-qa: Multimodal theory of mind question answering. In ACL, pages 16077–16102.
  19. John H Kagel and Dan Levin. 1986. The winner’s curse and public information in common value auctions. The American economic review. find
  20. Caixin Kang, Yifei Huang, Liangyang Ouyang, Mingfang Zhang, and Yoichi Sato. 2025. Can mllms read the room? a multimodal benchmark for verifying truthfulness in multi-party social interactions. arXiv preprint arXiv:2510.27195. — [arXiv:2510.27195](https://arxiv.org/abs/2510.27195) find
  21. Hyunwoo Kim, Melanie Sclar, Xuhui Zhou, Ronan Bras, Gunhee Kim, Yejin Choi, and Maarten Sap. 2023. Fantom: A benchmark for stress-testing machine theory of mind in interactions. In EMNLP, pages 14397–14413. find
  22. Katherine J Klein and Steve WJ Kozlowski. 2000. A multilevel approach to theory and research in organizations: Contextual, temporal, and emergent processes. Multilevel theory, research, and methods in organizations: Foundations, extensions, and new directions. find
  23. Kurt Lewin. 1951. Field theory in social science: selected theoretical papers (edited by dorwin cartwright.). find
  24. Huao Li, Yu Quan Chong, Simon Stepputtis, Joseph Campbell, Dana Hughes, Charles Lewis, and Katia P. Sycara. 2023. Theory of mind for multi-agent collaboration via large language models. In EMNLP, pages 180–192.
  25. Yuxuan Li, Vijay Veerabadran, Michael L Iuzzolino, Brett D Roads, Asli Celikyilmaz, and Karl Ridge-way. 2025. Egotom: Benchmarking theory of mind reasoning from egocentric videos. arXiv preprint arXiv:2503.22152. — [arXiv:2503.22152](https://arxiv.org/abs/2503.22152) find
  26. Chengzhi Liu, Zhongxing Xu, Qingyue Wei, Juncheng Wu, James Zou, Xin Eric Wang, Yuyin Zhou, and Sheng Liu. 2025a. More thinking, less seeing? assessing amplified hallucination in multimodal reasoning models. arXiv preprint arXiv:2505.21523. — [arXiv:2505.21523](https://arxiv.org/abs/2505.21523) find
  27. Yiwei Liu, Emma Jane Pretty, Jiahao Huang, and Saku Sugawara. 2025b. Tactfultom: Do llms have the theory of mind ability to understand white lies? In EMNLP, pages 25043–25061. find
  28. Andrei Lupu, Timon Willi, and Jakob N. Foerster. 2025. The decrypto benchmark for multi-agent reasoning and theory of mind. arXiv preprint arXiv:2506.20664. — [arXiv:2506.20664](https://arxiv.org/abs/2506.20664) find
  29. Leena Mathur, Marian Qian, Paul Pu Liang, and Louis-Philippe Morency. 2025. Social genome: Grounded social reasoning abilities of multimodal models. In EMNLP, pages 24868–24891. find
  30. Meta. 2024. Llama-3.2-11b-vision-instruct. Accessed: 2025-12-20. find
  31. Elisabeth Noelle-Neumann. 1974. The spiral of silence a theory of public opinion. Journal of communication, 24. find
  32. OpenAI. 2024. Gpt-4o system card. Accessed: 2025-12-20.
  33. OpenAI. 2025. Models: Gpt-5 series. Accessed: 2025-12-20. find
  34. OpenGVLab. 2025. Internvl3.5-8b. Accessed: 2025-12-20. find
  35. Joon Sung Park, Joseph C. O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. In UIST, pages 2:1–2:22. find
  36. David Premack and Guy Woodruff. 1978. Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences, 1. find
  37. Qwen. 2024. Qwen2-vl-7b-instruct. Accessed: 2025-12-20. find
  38. Qwen. 2025a. Qwen2.5-vl-7b-instruct. 2025-12-20. find
  39. Accessed: Qwen. 2025b. Qwen3-vl-8b-instruct. Accessed: 2025-12-20. find
  40. Maohao Ran, Zhenglin Wan, Cooper Lin, Yanting Zhang, Hongyu Xin, Hongwei Fan, Yibo Xu, Beier Luo, Yaxin Zhou, Wangbo Zhao, Lijie Yang, Lang Feng, Fuchao Yang, Jingxuan Wu, Yiqiao Huang, Chendong Ma, Dailing Jiang, Jianbo Deng, Sirui Han, and 4 others. 2026. Caveagent: Transforming llms into stateful… — [arXiv:2601.01569](https://arxiv.org/abs/2601.01569) find
  41. Matthew Riemer, Zahra Ashktorab, Djallel Bouneffouf, Payel Das, Miao Liu, Justin D. Weisz, and Murray Campbell. 2025. Position: Theory of mind benchmarks are broken for large language models. In ICML. find
  42. Maarten Sap, Ronan Le Bras, Daniel Fried, and Yejin Choi. 2022. Neural theory-of-mind? on the limits of social intelligence in large lms. In EMNLP, pages 3762–3780. find
  43. Haojun Shi, Suyu Ye, Xinyu Fang, Chuanyang Jin, Leyla Isik, Yen-Ling Kuo, and Tianmin Shu. 2025. Muma-tom: Multi-modal multi-agent theory of mind. In AAAI, pages 1510–1519. find
  44. Kazutoshi Shinoda, Nobukatsu Hojo, Kyosuke Nishida, Saki Mizuno, Keita Suzuki, Ryo Masumura, Hiroaki Sugiyama, and Kuniko Saito. 2025. Tomato: Verbalizing the mental states of role-playing llms for benchmarking theory of mind. In AAAI, pages 1520– 1528. find
  45. Xuhui Zhou, Hao Zhu, Leena Mathur, Ruohong Zhang, Haofei Yu, Zhengyang Qi, Louis-Philippe Morency, Yonatan Bisk, Daniel Fried, Graham Neubig, and Maarten Sap. 2024. Sotopia: Interactive evaluation for social intelligence in language agents. In ICLR.
  46. Emilio Villa-Cueva, SM Ahmed, Rendi Chevi, Jan Christian Blaise Cruz, Kareem Elzeky, Fermin Cristobal, Alham Fikri Aji, Skyler Wang, Rada Mihalcea, and Thamar Solorio. 2025. Moments: A comprehensive multimodal benchmark for theory of mind. arXiv preprint arXiv:2507.04415. find
  47. Zhenglin Wan, Xingrui Yu, David Mark Bossens, Yueming Lyu, Qing Guo, Flint Xiaofeng Fan, Yew Soon Ong, and Ivor Tsang. 2025. Diversifying policy behaviors with extrinsic behavioral curiosity. arXiv preprint arXiv:2410.06151. — [arXiv:2410.06151](https://arxiv.org/abs/2410.06151) find
  48. Qiaosi Wang, Xuhui Zhou, Maarten Sap, Jodi Forlizzi, and Hong Shen. 2025. Rethinking theory of mind benchmarks for llms: Towards a user-centered perspective. arXiv preprint arXiv:2504.10839. — [arXiv:2504.10839](https://arxiv.org/abs/2504.10839) find
  49. Zhiyuan Weng, Guikun Chen, and Wenguan Wang. 2025. Do as we do, not as you think: the conformity of large language models. In ICLR. find
  50. Yufan Wu, Yinghui He, Yilin Jia, Rada Mihalcea, Yulong Chen, and Naihao Deng. 2023. Hi-tom: A benchmark for evaluating higher-order theory of mind reasoning in large language models. In EMNLP, pages 10691–10706. find
  51. Hainiu Xu, Runcong Zhao, Lixing Zhu, Jinhua Du, and Yulan He. 2024. Opentom: A comprehensive benchmark for evaluating theory-of-mind reasoning capabilities of large language models. In ACL, pages 8593–8623. find
  52. Zixiang Xu, Yanbo Wang, Yue Huang, Jiayi Ye, Haomin Zhuang, Zirui Song, Lang Gao, Chenxi Wang, Zhaorun Chen, Yujun Zhou, Sixian Li, Wang Pan, Yue Zhao, Jieyu Zhao, Xiangliang Zhang, and Xiuying Chen. 2025. Socialmaze: A benchmark for evaluating social reasoning in large language models. arXiv prepri… find
  53. Shukang Yin, Chaoyou Fu, Sirui Zhao, Ke Li, Xing Sun, Tong Xu, and Enhong Chen. 2024. A survey on multimodal large language models. National Science Review, 11.
  54. Weihao Yu, Zhengyuan Yang, Linjie Li, Jianfeng Wang, Kevin Lin, Zicheng Liu, Xinchao Wang, and Lijuan Wang. 2023. Mm-vet: Evaluating large multimodal models for integrated capabilities. arXiv preprint arXiv:2308.02490.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tang-2026-grouptom-bench-benchmarking-group-theory/