AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

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

Haotong Sun, Jianye Xie, Bocheng Xu, Yinghui Jiang

ICML 2026 · Large language models › Reasoning and chain of thought · approach: Efficient and adaptive thinking

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes intrinsic dual-perspective reasoning for role-playing LLMs that realizes character-authenticity and user-satisfaction duality within the core attention mechanism rather than through prompts, auxiliary modules, or inference-time reflection.

Problem

Role-playing LLMs must balance character authenticity with user satisfaction; prior dual-process and dual-perspective approaches realize this duality externally (prompt conditioning, auxiliary modules, or inference-time reflection).

Approach

Realizes the dual perspectives (self/character and user) intrinsically inside the model's attention mechanism instead of externally via prompts, auxiliary modules, or reflection.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune base on same data and recipeQwen3-4B-Thinking · CharacterBench AVG (zh)average score: 3.69 -> 3.80Improved
add KSKT architecture over data-matched fine-tuning (ablation)Qwen3-4B-Thinking + FT · CharacterBench AVG (zh)average score: 3.80 -> 3.92Improved
use wider unified attention with proxy losses instead of dual-stream factorization (ablation)KSKT · CharacterBench AVG (zh)average score: 3.83 vs 3.92Hurt
add Dual-Stream Axial Attention alone (ablation)Qwen3-4B-Thinking · CharacterBench Emotiondelta over base: +0.39Improved
add Mutual-Understanding Position Encoding alone (ablation)Qwen3-4B-Thinking · CharacterBench Emotiondelta over base: +0.41Improved
add Self-Awareness MoE alone (ablation)Qwen3-4B-Thinking · CharacterBench Personadelta over base: +0.28Improved
add Bipolar Reasoning Module alone (ablation)Qwen3-4B-Thinking · CharacterBench Believability and Knowledgedelta over base: +0.18 eachImproved
add Self-Awareness MoE alone (ablation)Qwen3-4B-Thinking · CharacterBench Memorydelta over base: -0.05Hurt
add KSKT to base modelQwen3-4B-Thinking · SOTOPIA RelationshipRelationship score: 1.45 -> 1.73Improved
add KSKT to base modelQwen3-4B-Thinking · SOTOPIA SecretSecret score: -0.07 -> -0.18Hurt
add 3-shot prompting (ablation)Qwen3-4B-Thinking · SOTOPIA OverallOverall score: 3.08 -> 3.28Improved
add 3-shot prompting (ablation)KSKT · SOTOPIA OverallOverall score: 3.30 -> 3.48Improved

Experimental setup

Builds on: dual-process theory dual-perspective reasoning

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy