Elamur Research⌕
☰
home / papers / 2026 / PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints

PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints

Paper · 2026 · Minjun Park, Donghyun Kim, Hyeonjong Ju et al.

Summary

As people and organizations deploy their own AI agents, those agents will need to cooperate without exposing their owners' confidential information. Current multi-agent benchmarks measure task completion and coordination but do not model privacy. PAC-BENCH (Private Agent Collaboration Benchmark), from Yonsei University, addresses this gap. Each scenario pairs two agents serving different owners. Every agent has a profile, a private memory and natural-language privacy constraints, and both share a goal that requires producing a concrete artifact such as a table, document or database. Scenarios come from a multi-stage LLM generation pipeline seeded with industry domains from the MSCI classification, followed by quality filtering with LLM-as-a-Judge and human annotators. Agents act turn by turn, and the resulting trajectory is scored separately for task success and privacy leakage. Experiments show that privacy constraints sharply reduce joint performance. Results also depend more on the agent that starts the conversation than on its partner. The authors trace this drop to recurring coordination failures and argue that privacy-aware collaboration is an unresolved problem needing new coordination mechanisms.

Key points

  • Dataset size: a core set of 100 human-validated two-agent scenarios plus an extended set of 1,476 scenarios. Scenarios span 74 industry domains drawn from the MSCI standard classification.
  • Evaluation: collaborations run for up to 20 turns. Two scores are reported:
    • a Task Score (TS), computed against goal requirements;
    • a Privacy Score (PS), computed by keyword filtering followed by rubric-based G-Eval.
  • Main findings: privacy constraints substantially degrade collaboration, and the initiating agent dominates the outcome.
  • Failure modes identified:
    • privacy leaks early in the dialogue;
    • abstraction so cautious that the information becomes useless;
    • hallucinations induced by withholding information.
  • Data and limitations: all data is synthetic, with no real PII, and code is released at github.com/PAC-Bench/PAC-Bench. Stated limitations include the two-owner setup only, fixed constraints and prompting, and reliance on an LLM judge.

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

🎓Yonsei UniversitySeoul · South Korea17 works · 501 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 works

Cite

Cite as Minjun Park, Donghyun Kim, Hyeonjong Ju, Seungwon Lim, Dongwook Choi, Taeyoon Kwon, Minju Kim and Jinyoung Yeo. 2026. PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1552

@inproceedings{park2026pac,
  title = {{PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints}},
  author = {Minjun Park, Donghyun Kim, Hyeonjong Ju, Seungwon Lim, Dongwook Choi, Taeyoon Kwon, Minju Kim and Jinyoung Yeo},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1552},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1552}
}

References

53 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. 2024. Information technology — security techniques — privacy framework. find
  2. A2A Protocol. 2025. What is a2a? Accessed: 2025-12-23. find
  3. Anthropic. 2025. Claude 3.7: Model overview. Accessed: 2025-12-23. find
  4. Anthropic and 1 others. 2024. Model context protocol. https://github.com/ modelcontextprotocol. GitHub repository. ↗
  5. Hyungjoo Chae, Sunghwan Kim, Junhee Cho, Seungone Kim, Seungjun Moon, Gyeom Hwangbo, Dongha Lim, Minjin Kim, Yeonjun Hwang, Minju Gwak, and 1 others. 2025. Web-shepherd: Advancing prms for reinforcing web agents. Advances in Neural Information Processing Systems. find
  6. Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chi-Min Chan, Heyang Yu, Yaxi Lu, Yi-Hsin Hung, Chen Qian, and 1 others. 2024. Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors. In ICLR.
  7. Haonan Duan, Adam Dziedzic, Nicolas Papernot, and Franziska Boenisch. 2023. Flocks of stochastic parrots: Differentially private prompt learning for large language models. Advances in Neural Information Processing Systems, 36:76852–76871. find
  8. Adam Fourney, Gagan Bansal, Hussein Mozannar, Cheng Tan, Eduardo Salinas, Friederike Niedtner, Grace Proebsting, Griffin Bassman, Jack Gerrits, Jacob Alber, and 1 others. 2024. Magentic-one: A generalist multi-agent system for solving complex tasks. arXiv preprint arXiv:2411.04468. find
  9. Longling Geng and Edward Y. Chang. 2025. Realm-bench: A benchmark for evaluating multi-agent systems on real-world, dynamic planning and scheduling tasks. Preprint, arXiv:2502.18836. — [arXiv:2502.18836](https://arxiv.org/abs/2502.18836) find
  10. Amr Gomaa, Ahmed Salem, and Sahar Abdelnabi. 2025. Converse: Benchmarking contextual safety in agent-to-agent conversations. arXiv preprint arXiv:2511.05359. find
  11. Diego Gosmar and Deborah A. Dahl. 2025. Sentinel agents for secure and trustworthy agentic ai in multi-agent systems. arXiv preprint arXiv:2509.14956. — [arXiv:2509.14956](https://arxiv.org/abs/2509.14956) find
  12. Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V Chawla, Olaf Wiest, and Xiangliang Zhang. 2024. Large language model based multi-agents: A survey of progress and challenges. arXiv preprint arXiv:2402.01680.
  13. Shanshan Han, Qifan Zhang, Yuhang Yao, Weizhao Jin, and Zhaozhuo Xu. 2024. Llm multi-agent systems: Challenges and open problems. arXiv preprint arXiv:2402.03578. — [arXiv:2402.03578](https://arxiv.org/abs/2402.03578) find
  14. Junda He, Christoph Treude, and David Lo. 2025a. Llm-based multi-agent systems for software engineering: Literature review, vision, and the road ahead. ACM Transactions on Software Engineering and Methodology, 34(5):1–30. find
  15. Pengfei He, Yuping Lin, Shen Dong, Han Xu, Yue Xing, and Hui Liu. 2025b. Red-teaming llm multi-agent systems via communication attacks. In Findings of the Association for Computational Linguistics: ACL 2025, pages 6726–6747. find
  16. Yubo Dong, Xukun Zhu, Zhengzhe Pan, Linchao Zhu, and Yi Yang. 2024. Villageragent: A graph-based multi-agent framework for coordinating complex task dependencies in minecraft. arXiv preprint arXiv:2406.05720. find
  17. Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, and 1 others. 2023. Metagpt: Meta programming for a multi-agent collaborative framework. In The Twelfth International Conference on Learning Representations.
  18. Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. 2023. Improving factuality and reasoning in language models through multiagent debate. In Forty-first International Conference on Machine Learning.
  19. Wenyue Hua, Lizhou Fan, Lingyao Li, Kai Mei, Jianchao Ji, Yingqiang Ge, Libby Hemphill, and Yongfeng Zhang. 2023. War and peace (waragent): Large language model-based multi-agent simulation of world wars. arXiv preprint arXiv:2311.17227. find
  20. Yoichi Ishibashi and Yoshimasa Nishimura. 2024. Self-organized agents: A llm multi-agent framework toward ultra large-scale code generation and optimization. arXiv preprint arXiv:2404.02183. — [arXiv:2404.02183](https://arxiv.org/abs/2404.02183) find
  21. Hannah Rose Kirk, Bertie Vidgen, Paul Röttger, and Scott A Hale. 2024. The benefits, risks and bounds of personalizing the alignment of large language models to individuals. Nature Machine Intelligence, 6(4):383–392. find
  22. Taeyoon Kwon, Dongwook Choi, Hyojun Kim, Sunghwan Kim, Seungjun Moon, Beong-woo Kwak, Kuan-Hao Huang, and Jinyoung Yeo. 2025. Embodied agents meet personalization: Investigating challenges and solutions through the lens of memory utilization. arXiv preprint arXiv:2505.16348. — [arXiv:2505.16348](https://arxiv.org/abs/2505.16348) find
  23. Jisoo Lee, Raeyoung Chang, Dongwook Kwon, Harmanpreet Singh, and Nikhil Verma. 2025. Gemmas: Graph-based evaluation metrics for multi agent systems. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 1522–1532. find
  24. Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. 2023. Camel: Communicative agents for" mind" exploration of large language model society. Advances in Neural Information Processing Systems, 36:51991–52008.
  25. Hao Li, Chenghao Yang, An Zhang, Yang Deng, Xiang Wang, and Tat-Seng Chua. 2025. Hello again! llm-powered personalized agent for long-term dialogue. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technolog…
  26. Yuanchun Li, Hao Wen, Weijun Wang, Xiangyu Li, Yizhen Yuan, Guohong Liu, Jiacheng Liu, Wenxing Xu, Xiang Wang, Yi Sun, and 1 others. 2024. Personal llm agents: Insights and survey about the capability, efficiency and security. arXiv preprint arXiv:2401.05459. find
  27. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-eval: Nlg evaluation using gpt-4 with better human alignment. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 2511–2522.
  28. Meta AI. 2024. LLaMA 3.3: Open foundation and fine-tuned language models. Accessed: 2025-12-23. find
  29. Michele Miranda, Elena Sofia Ruzzetti, Andrea Santilli, Fabio Massimo Zanzotto, Sébastien Bratières, and Emanuele Rodolà. 2024. Preserving privacy in large language models: A survey on current threats and solutions. CoRR, abs/2408.05212. — [arXiv:2408.05212](https://arxiv.org/abs/2408.05212) find
  30. Mistral. 2025. Introducing mistral 3. Accessed: 2025-12-23. find
  31. MSCI Inc. 2025. Global industry classification standard (gics®). Accessed: 2025-12-23. find
  32. Mason Nakamura, Abhinav Kumar, Saaduddin Mahmud, Sahar Abdelnabi, Shlomo Zilberstein, and Eugene Bagdasarian. 2025. Terrarium: Revisiting the blackboard for multi-agent safety, privacy, and security studies. arXiv preprint arXiv:2510.14312. — [arXiv:2510.14312](https://arxiv.org/abs/2510.14312) find
  33. OpenAI. 2025. GPT-5.1: Technical overview. Accessed: 2025-12-23. find
  34. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th annual acm symposium on user interface software and technology, pages 1–22. find
  35. Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, and 1 others. 2024a. Chatdev: Communicative agents for software development. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pa…
  36. Chen Qian, Zihao Xie, Yifei Wang, Wei Liu, Kunlun Zhu, Hanchen Xia, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, and 1 others. 2024b. Scaling large language model-based multi-agent collaboration. arXiv preprint arXiv:2406.07155. find
  37. Chris Richardson, Yao Zhang, Kellen Gillespie, Sudipta Kar, Arshdeep Singh, Zeynab Raeesy, Omar Zia Khan, and Abhinav Sethy. 2023. Integrating summarization and retrieval for enhanced personalization via large language models. arXiv preprint arXiv:2310.20081. — [arXiv:2310.20081](https://arxiv.org/abs/2310.20081) find
  38. Alireza Salemi, Surya Kallumadi, and Hamed Zamani. 2024. Optimization methods for personalizing large language models through retrieval augmentation. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 752–762. find
  39. Robin Staab, Mark Vero, Mislav Balunović, and Martin Vechev. 2023. Beyond memorization: Violating privacy via inference with large language models. arXiv preprint arXiv:2310.07298. — [arXiv:2310.07298](https://arxiv.org/abs/2310.07298) find
  40. Xinyu Tang, Richard Shin, Huseyin A Inan, Andre Manoel, Fatemehsadat Mireshghallah, Zinan Lin, Sivakanth Gopi, Janardhan Kulkarni, and Robert Sim. 2024. Privacy-preserving in-context learning with differentially private few-shot generation. In ICLR. find
  41. Qwen Team. 2025. Qwen3 technical report. Preprint, arXiv:2505.09388.
  42. Khanh-Tung Tran, Dung Dao, Minh-Duong Nguyen, Quoc-Viet Pham, Barry O’Sullivan, and Hoang D Nguyen. 2025. Multi-agent collaboration mechanisms: A survey of llms. arXiv preprint arXiv:2501.06322. find
  43. Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. 2023. Voyager: An open-ended embodied agent with large language models. arXiv preprint arXiv:2305.16291. find
  44. Wei Wang, Dan Zhang, Tao Feng, Boyan Wang, and Jie Tang. 2024. Battleagentbench: A benchmark for evaluating cooperation and competition capabilities of language models in multi-agent systems. arXiv preprint arXiv:2408.15971. — [arXiv:2408.15971](https://arxiv.org/abs/2408.15971) find
  45. Oshadha Wijerathne, Amandi Nimasha, Dushan Fernando, Nisansa de Silva, and Srinath Perera. 2025. Scheduleme: Multi-agent calendar assistant. arXiv preprint arXiv:2509.25693. find
  46. Hui Yang, Sifu Yue, and Yunzhong He. 2023. Auto-gpt for online decision making: Benchmarks and additional opinions. arXiv preprint, arXiv:2306.02224. find
  47. Ziyi Yang, Zaibin Zhang, Zirui Zheng, Yuxian Jiang, Ziyue Gan, Zhiyu Wang, Zijian Ling, Martz Ma, Bowen Dong, Prateek Gupta, and 1 others. Oasis: Open agents social interaction simulations on a large scale. find
  48. Yifan Yao, Jinhao Duan, Kaidi Xu, Yuanfang Cai, Zhibo Sun, and Yue Zhang. 2024. A survey on large language model (llm) security and privacy: The good, the bad, and the ugly. High-Confidence Computing, 4(2):100211.
  49. Weizhi Zhang, Xinyang Zhang, Chenwei Zhang, Liangwei Yang, Jingbo Shang, Zhepei Wei, Henry Peng Zou, Zijie Huang, Zhengyang Wang, Yifan Gao, Xiaoman Pan, Lian Xiong, Jingguo Liu, Philip S. Yu, and Xian Li. 2025. Personaagent: When large language model agents meet personalization at test time. Prepri… find
  50. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, and 1 others. 2023. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in neural information processing systems, 36:46595–46623.
  51. Shuyan Zhou, Frank F Xu, Hao Zhu, Xuhui Zhou, Robert Lo, Abishek Sridhar, Xianyi Cheng, Yonatan Bisk, Daniel Fried, Uri Alon, and 1 others. 2023a. Webarena: A realistic web environment for building autonomous agents. arXiv preprint arXiv:2307.13854.
  52. Wangchunshu Zhou, Yuchen Eleanor Jiang, Long Li, Jialong Wu, Tiannan Wang, Shi Qiu, Jintian Zhang, Jing Chen, Ruipu Wu, Shuai Wang, Shiding Zhu, Jiyu Chen, Wentao Zhang, Xiangru Tang, Ningyu Zhang, Huajun Chen, Peng Cui, and Mrinmaya Sachan. 2023b. Agents: An open-source framework for autonomous lan… find
  53. Kunlun Zhu, Hongyi Du, Zhaochen Hong, Xiaocheng Yang, Shuyi Guo, Daisy Zhe Wang, Zhenhailong Wang, Cheng Qian, Robert Tang, Heng Ji, and 1 others. 2025. Multiagentbench: Evaluating the collaboration and competition of llm agents. In Proceedings of the 63rd Annual Meeting of the Association for Compu… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/park-2026-pac-bench-evaluating-multi-agent/