Know Thyself, Know Thy User: Intrinsic Dual-Perspective Reasoning for Role-Playing LLMs

Haotong Sun, Jianye Xie, Bocheng Xu, Yinghui Jiang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:117351-117383, 2026.

Abstract

As role-playing Large Language Models (LLMs) become central to personalized AI, they face a fundamental challenge: balancing character authenticity with user satisfaction. Prior dual-process and dual-perspective approaches address this through prompt-level conditioning, auxiliary modules, or inference-time reflection—realizing duality externally rather than within the core attention mechanism. We introduce the KnowSelf-KnowOther Transformer (KSKT), which embeds dual-perspective reasoning directly into the generation process via dual-stream factorized attention that processes self-understanding and other-understanding through separate streams. This intrinsic integration enables token-level dynamic balance rather than post-hoc reconciliation. On CharacterBench, KSKT achieves 6.2% overall improvement over Qwen3-4B-Thinking. On SOTOPIA, KSKT improves Relationship by 19.3% over the base model—the dimension requiring explicit self-other coordination. These results establish intrinsic dual-perspective reasoning as an effective architectural principle for role-playing systems.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-sun26at, title = {Know Thyself, Know Thy User: Intrinsic Dual-Perspective Reasoning for Role-Playing {LLM}s}, author = {Sun, Haotong and Xie, Jianye and Xu, Bocheng and Jiang, Yinghui}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {117351--117383}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/sun26at/sun26at.pdf}, url = {https://proceedings.mlr.press/v306/sun26at.html}, abstract = {As role-playing Large Language Models (LLMs) become central to personalized AI, they face a fundamental challenge: balancing character authenticity with user satisfaction. Prior dual-process and dual-perspective approaches address this through prompt-level conditioning, auxiliary modules, or inference-time reflection—realizing duality externally rather than within the core attention mechanism. We introduce the KnowSelf-KnowOther Transformer (KSKT), which embeds dual-perspective reasoning directly into the generation process via dual-stream factorized attention that processes self-understanding and other-understanding through separate streams. This intrinsic integration enables token-level dynamic balance rather than post-hoc reconciliation. On CharacterBench, KSKT achieves 6.2% overall improvement over Qwen3-4B-Thinking. On SOTOPIA, KSKT improves Relationship by 19.3% over the base model—the dimension requiring explicit self-other coordination. These results establish intrinsic dual-perspective reasoning as an effective architectural principle for role-playing systems.} }
Endnote
%0 Conference Paper %T Know Thyself, Know Thy User: Intrinsic Dual-Perspective Reasoning for Role-Playing LLMs %A Haotong Sun %A Jianye Xie %A Bocheng Xu %A Yinghui Jiang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-sun26at %I PMLR %P 117351--117383 %U https://proceedings.mlr.press/v306/sun26at.html %V 306 %X As role-playing Large Language Models (LLMs) become central to personalized AI, they face a fundamental challenge: balancing character authenticity with user satisfaction. Prior dual-process and dual-perspective approaches address this through prompt-level conditioning, auxiliary modules, or inference-time reflection—realizing duality externally rather than within the core attention mechanism. We introduce the KnowSelf-KnowOther Transformer (KSKT), which embeds dual-perspective reasoning directly into the generation process via dual-stream factorized attention that processes self-understanding and other-understanding through separate streams. This intrinsic integration enables token-level dynamic balance rather than post-hoc reconciliation. On CharacterBench, KSKT achieves 6.2% overall improvement over Qwen3-4B-Thinking. On SOTOPIA, KSKT improves Relationship by 19.3% over the base model—the dimension requiring explicit self-other coordination. These results establish intrinsic dual-perspective reasoning as an effective architectural principle for role-playing systems.
APA
Sun, H., Xie, J., Xu, B. & Jiang, Y.. (2026). Know Thyself, Know Thy User: Intrinsic Dual-Perspective Reasoning for Role-Playing LLMs. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:117351-117383 Available from https://proceedings.mlr.press/v306/sun26at.html.

Related Material