Simulating Before Planning: Constructing Intrinsic User World Model for User-Tailored Dialogue Policy Planning

SIGIR 25 2025 50.7 method

TLDR

Proposes UDP framework with Intrinsic User World Model for user-tailored dialogue policy planning using diffusion and Brownian Bridge.

Reasoning

Strengths include addressing the overlooked role of user characteristics in dialogue policy planning and introducing a novel three-stage framework. Weaknesses are the lack of real-world experimental validation and limited generalizability beyond dialogue tasks.

Read-first score

Read-first score 50.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 31.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,experiment,result

Topical relevance 42%

44.3

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 366.

Keyword Scores

world model

generative world model

world simulator

interactive world model

model-based reinforcement learning world model

world dynamics prediction

video world model

Deep Analysis

Innovations

Intrinsic User World Model for user-tailored dialogue policy planning
Three-stage UDP framework: User Persona Portraying (diffusion model), User Feedback Anticipating (Brownian Bridge-inspired anticipator), User-Tailored Policy Planning
Active learning approach that prioritizes challenging user personas during training

Methodology

The paper first conducts a comprehensive study using task-specific user personas to assess existing dialogue policy planning approaches, revealing their limitations. Then it proposes the UDP framework, which incorporates an Intrinsic User World Model. UDP operates in three stages: (1) User Persona Portraying using a diffusion model to dynamically infer user profiles; (2) User Feedback Anticipating using a Brownian Bridge-inspired anticipator to predict user reactions; (3) User-Tailored Policy Planning integrating these insights to optimize response strategies. An active learning approach is further proposed to prioritize challenging user personas during training. Experiments are conducted on benchmarks with collaborative and non-collaborative settings.

Key Results

Comprehensive experiments demonstrate the effectiveness of UDP in learning user-specific dialogue strategies, validating the protocol's utility and highlighting UDP's robustness and adaptability.

Resources

Paper