--- title: "From Propositional to Perceptual Asymmetry — Zhu 2026" description: "Frictive Policy Optimization (FPO) treats misunderstanding and repair in collaborative dialogue as useful epistemic evidence rather than noise." authors: ["Yifan Zhu", "Kyeongmin Rim", "James Pustejovsky"] year: 2026 venue: "Proceedings of the 27th Annual Meeting of the Special Interest Group on Discourse and Dialogue" source: "https://aclanthology.org/2026.sigdial-1.28/" doi: "10.18653/v1/2026.sigdial-1.28" 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/zhu-2026-propositional-perceptual-asymmetry-extending-fpo/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhu-2026-propositional-perceptual-asymmetry-extending-fpo/) [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/5/) / From Propositional to Perceptual Asymmetry: Extending FPO to Asymmetric Partial Information Dialogue # From Propositional to Perceptual Asymmetry: Extending FPO to Asymmetric Partial Information Dialogue **Authors:** [Yifan Zhu](https://md.research.elamur.ai/free/en/authors/yifan-zhu/), [Kyeongmin Rim](https://md.research.elamur.ai/free/en/authors/kyeongmin-rim/), [James Pustejovsky](https://md.research.elamur.ai/free/en/authors/james-pustejovsky/) · **Year:** 2026 · **Venue:** Proceedings of the 27th Annual Meeting of the Special Interest Group on Discourse and Dialogue · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2026.sigdial-1.28/) · [PDF](https://aclanthology.org/2026.sigdial-1.28.pdf) · [DOI 10.18653/v1/2026.sigdial-1.28](https://doi.org/10.18653/v1/2026.sigdial-1.28) · [OpenAlex](https://openalex.org/W7214041997) **Citations:** 0 citations (openalex-snapshot · 2026-10-14) ## Summary Frictive Policy Optimization (FPO) treats misunderstanding and repair in collaborative dialogue as useful epistemic evidence rather than noise. Its existing implementation, FAAF, assumes participants see the same scene and differ only in beliefs about it. The authors extend FPO to perceptual asymmetry, where each participant holds a different partial view, so one referring expression can point to different referents. They reformulate the friction functional to be computed from within each participant's own information horizon. In this formulation, speakers incur friction from ambiguity in their own frame, and addressees incur it from conflicts across frames. The extension is tested on the HCRC MapTask by joining NXT dialogue-act labels with the per-perspective GMMT annotation. The authors also run LLM probes and compare against OneCommon. A few ambiguous multiplicity configurations cause a disproportionate share of misunderstandings, often through chains that seem successful but quietly diverge. Perspective-bound probing beats omniscient access to both maps, even for a frontier model. This suggests that adding context or scaling models does not solve grounding under private information. ## Key points - **Data:** The study uses 128 HCRC MapTask dialogues, joining the NXT release (v2.1) with GMMT (Li et al., 2025). GMMT covers 13,077 reference expressions labeled aligned, pending or misunderstood. - **Multiplicity discrepancies:** These make up about 6% of landmarks, yet their misunderstanding rate is 12.0%, against a 1.8% corpus average. Multiplicity reference chains show a 38.9% decay rate. - **LLM probing:** The perspectival setting detected all 13 gold repair events with both Gemma-4-31B and GPT-5. The omniscient setting found only 5 and 12 respectively, and the detection-to-repair rate was 0.302. - **OneCommon comparison (5,191 dialogues):** The selection failure rate is 56%. Dialogues containing ambiguous markables need more turns (5.64 vs. 4.53). Surface signals such as hedging do not predict task success. - **Proposed annotation refinements:** - split the pending state into subtypes; - separate negotiated alignment from silently accommodated alignment by linking reference trajectories to dialogue-act structure. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/) ## Cite The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{zhu2026propositional, title = {{From Propositional to Perceptual Asymmetry: Extending FPO to Asymmetric Partial Information Dialogue}}, author = {Zhu, Yifan and Rim, Kyeongmin and Pustejovsky, James}, booktitle = {Proceedings of the 27th Annual Meeting of the Special Interest Group on Discourse and Dialogue}, year = {2026}, doi = {10.18653/v1/2026.sigdial-1.28}, url = {https://doi.org/10.18653/v1/2026.sigdial-1.28} } ``` - [← previous From Noise to Signal: When Outliers Seed New Topics](https://md.research.elamur.ai/free/en/papers/zve-2026-noise-signal-when-outliers-seed/) - [next → From Rosetta to Match-Up: A Paired Corpus of Linguistic Puzzles with Human and LLM Benchmarks](https://md.research.elamur.ai/free/en/papers/majmudar-2026-rosetta-match-up-paired-corpus/) - [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/5/) ## References 14 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Anne H. Anderson, Miles Bader, Ellen Gurman Bard, Elizabeth Boyle, Gwyneth Doherty, Simon Garrod, Stephen Isard, Jacqueline Kowtko, Jan McAllister, Jim Miller, Catherine Sotillo, Henry S. Thompson, and Regina Weinert. 1991. The HCRC Map Task corpus. Language and Speech, 34(4):351–366. [find](https://scholar.google.com/scholar?q=Anne+H%2E+Anderson%2C+Miles+Bader%2C+Ellen+Gurman+Bard%2C+Elizabeth+Boyle%2C+Gwyneth+Doherty%2C+Simon+Garrod%2C+Stephen+Isard%2C+Jacqueline+Kowtko%2C+Jan+McAllister%2C+Jim+Miller%2C+Catherine+Sotillo%2C+Henry+S%2E+Thompson%2C+an) 2. Paul F. Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. 2017. Deep reinforcement learning from human preferences. Advances in Neural Information Processing Systems, 30. [find](https://scholar.google.com/scholar?q=Paul+F%2E+Christiano%2C+Jan+Leike%2C+Tom+Brown%2C+Miljan+Martic%2C+Shane+Legg%2C+and+Dario+Amodei%2E+2017%2E+Deep+reinforcement+learning+from+human+preferences%2E+Advances+in+Neural+Information+Processing+Systems%2C+30%2E) 3. Herbert H. Clark. 1996. Using Language. Cambridge University Press. [find](https://scholar.google.com/scholar?q=Herbert+H%2E+Clark%2E+1996%2E+Using+Language%2E+Cambridge+University+Press%2E) 4. Herbert H. Clark and Susan E. Brennan. 1991. Grounding in communication. In Lauren B. Resnick, John M. Levine, and Stephanie D. Teasley, editors, Perspectives on Socially Shared Cognition, pages 127–149. American Psychological Association. [find](https://scholar.google.com/scholar?q=Herbert+H%2E+Clark+and+Susan+E%2E+Brennan%2E+1991%2E+Grounding+in+communication%2E+In+Lauren+B%2E+Resnick%2C+John+M%2E+Levine%2C+and+Stephanie+D%2E+Teasley%2C+editors%2C+Perspectives+on+Socially+Shared+Cognition%2C+pages+127%E2%80%931) 5. Georgi Karadzhov, Tom Stafford, and Andreas Vlachos. 2023. DeliData: A dataset for deliberation in multi-party problem solving. Proceedings of the ACM on Human-Computer Interaction, 7(CSCW2):1–25. [find](https://scholar.google.com/scholar?q=Georgi+Karadzhov%2C+Tom+Stafford%2C+and+Andreas+Vlachos%2E+2023%2E+DeliData%3A+A+dataset+for+deliberation+in+multi%2Dparty+problem+solving%2E+Proceedings+of+the+ACM+on+Human%2DComputer+Interaction%2C+7%28CSCW2%29%3A1%E2%80%9325%2E) 6. Ibrahim Khalil Khebour, Mariah Bradford, Kenneth Lai, Christopher Tam, Jingxuan Tu, Benjamin A. Ibarra, Nathaniel Blanchard, Nikhil Krishnaswamy, and James Pustejovsky. 2024. Common ground tracking in multimodal dialogue. In Proceedings of the 2024 Joint International Conference on Computational Lin… [find](https://scholar.google.com/scholar?q=Ibrahim+Khalil+Khebour%2C+Mariah+Bradford%2C+Kenneth+Lai%2C+Christopher+Tam%2C+Jingxuan+Tu%2C+Benjamin+A%2E+Ibarra%2C+Nathaniel+Blanchard%2C+Nikhil+Krishnaswamy%2C+and+James+Pustejovsky%2E+2024%2E+Common+ground+tracking+in) [our card](https://md.research.elamur.ai/free/en/papers/khebour-2024-common-ground-tracking-multimodal-dialogue/) 7. Nan Li, Albert Gatt, and Massimo Poesio. 2025. Grounded misunderstandings in asymmetric dialogue: A perspectivist annotation scheme for Map-Task. arXiv preprint arXiv:2511.03718. To appear in LREC 2026. — [arXiv:2511.03718](https://arxiv.org/abs/2511.03718) [find](https://scholar.google.com/scholar?q=Nan+Li%2C+Albert+Gatt%2C+and+Massimo+Poesio%2E+2025%2E+Grounded+misunderstandings+in+asymmetric+dialogue%3A+A+perspectivist+annotation+scheme+for+Map%2DTask%2E+arXiv+preprint+arXiv%3A2511%2E03718%2E+To+appear+in+LREC+202) 8. Abhijnan Nath, Carine Graff, Andrei Bachinin, and Nikhil Krishnaswamy. 2025. Frictional agent alignment framework: Slow down and don’t break things. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11042–11089, Vienna, Austria.… [find](https://scholar.google.com/scholar?q=Abhijnan+Nath%2C+Carine+Graff%2C+Andrei+Bachinin%2C+and+Nikhil+Krishnaswamy%2E+2025%2E+Frictional+agent+alignment+framework%3A+Slow+down+and+don%E2%80%99t+break+things%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+the+As) 9. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D. Manning, Stefano Ermon, and Chelsea Finn. 2024. Direct preference optimization: Your language model is secretly a reward model. In Advances in Neural Information Processing Systems, volume 36. [find](https://scholar.google.com/scholar?q=Rafael+Rafailov%2C+Archit+Sharma%2C+Eric+Mitchell%2C+Christopher+D%2E+Manning%2C+Stefano+Ermon%2C+and+Chelsea+Finn%2E+2024%2E+Direct+preference+optimization%3A+Your+language+model+is+secretly+a+reward+model%2E+In+Advance) 10. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y.K. Li, Y. Wu, and 1 others. 2024. DeepSeekMath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300. [find](https://scholar.google.com/scholar?q=Zhihong+Shao%2C+Peiyi+Wang%2C+Qihao+Zhu%2C+Runxin+Xu%2C+Junxiao+Song%2C+Xiao+Bi%2C+Haowei+Zhang%2C+Mingchuan+Zhang%2C+Y%2EK%2E+Li%2C+Y%2E+Wu%2C+and+1+others%2E+2024%2E+DeepSeekMath%3A+Pushing+the+limits+of+mathematical+reasoning+in+) [our card](https://md.research.elamur.ai/free/en/papers/shao-2024-deepseekmath-pushing-limits-mathematical-reasoning/) 11. Aaditya Singh, Adam Fry, Adam Perelman, Adam Tart, Adi Ganesh, Ahmed El-Kishky, Aidan McLaughlin, Aiden Low, AJ Ostrow, Akhila Ananthram, Akshay Nathan, Alan Luo, Alec Helyar, Aleksander Madry, Aleksandr Efremov, Aleksandra Spyra, Alex Baker-Whitcomb, Alex Beutel, Alex Karpenko, and 465 others. 2025… [find](https://scholar.google.com/scholar?q=Aaditya+Singh%2C+Adam+Fry%2C+Adam+Perelman%2C+Adam+Tart%2C+Adi+Ganesh%2C+Ahmed+El%2DKishky%2C+Aidan+McLaughlin%2C+Aiden+Low%2C+AJ+Ostrow%2C+Akhila+Ananthram%2C+Akshay+Nathan%2C+Alan+Luo%2C+Alec+Helyar%2C+Aleksander+Madry%2C+Aleksa) [our card](https://md.research.elamur.ai/free/en/papers/singh-2025-openai-gpt-5-system-card/) 12. Gemma Team and Google DeepMind. 2026. Gemma 4: Open multimodal models based on gemini 3 technology. [find](https://scholar.google.com/scholar?q=Gemma+Team+and+Google+DeepMind%2E+2026%2E+Gemma+4%3A+Open+multimodal+models+based+on+gemini+3+technology%2E) 13. Takuma Udagawa and Akiko Aizawa. 2019. A natural language corpus of common grounding under continuous and partially-observable context. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, pages 7120–7127. [find](https://scholar.google.com/scholar?q=Takuma+Udagawa+and+Akiko+Aizawa%2E+2019%2E+A+natural+language+corpus+of+common+grounding+under+continuous+and+partially%2Dobservable+context%2E+In+Proceedings+of+the+AAAI+Conference+on+Artificial+Intelligence) 14. Yifan Zhu, Mariah Bradford, Kenneth Lai, Timothy Obiso, Videep Venkatesha, James Pustejovsky, and Nikhil Krishnaswamy. 2026. Distributed partial information puzzles: Examining common ground construction under epistemic asymmetry. arXiv preprint arXiv:2603.05450. [find](https://scholar.google.com/scholar?q=Yifan+Zhu%2C+Mariah+Bradford%2C+Kenneth+Lai%2C+Timothy+Obiso%2C+Videep+Venkatesha%2C+James+Pustejovsky%2C+and+Nikhil+Krishnaswamy%2E+2026%2E+Distributed+partial+information+puzzles%3A+Examining+common+ground+constructi)