Intent Speaks Louder: Controllable User Simulation Beyond Response Imitation

2026-08-10

AI Summary

Existing user simulators struggle to precisely control turn-level intents, often leading to deviations from the intended dialogue goals. This work proposes decoupling user intent from language generation by introducing, for the first time, an explicit turn-level intent interface that enables fine-grained control over user behavior through instruction-conditioned generation. Built upon the UserIDA framework, the approach integrates supervised fine-tuning with population-based reinforcement learning to calibrate intents effectively, substantially enhancing the simulator’s goal-directed capabilities. Experimental results demonstrate that the proposed method achieves an intent accuracy of 86.6%, representing a 24.3-percentage-point improvement over the strongest baseline, and successfully fulfills at least four distinct goal intents in 91.7% of dialogue states.

Abstract

User simulators are widely used as scalable environments for training and evaluating interactive assistants. Generating the next user turn is inherently one-to-many: the same profile and dialogue context may support multiple plausible continuations with different local interaction intents. A fluent response may therefore advance the dialogue through an inappropriate intent, such as acceptance rather than repair. Our key insight is that controllable user simulation should separate which local interaction intent the next user turn should realize from how that intent is expressed in language. We introduce UserIDA (User Intent-Directive Alignment), which exposes interaction intent as an explicit per-turn directive. UserIDA defines a six-way intent interface, learns directive-conditioned generation through supervised fine-tuning, and uses intent-calibrated policy optimization during group-based reinforcement learning. The reward preserves composite response quality while ensuring that intent-violating candidates rank below compliant alternatives in mixed groups. On LMSYS-USP, UserIDA achieves 86.6\% intent accuracy, outperforming the strongest dedicated user-simulator baseline by 24.3 percentage points while improving semantic and stylistic similarity. In within-context interventions, it realizes at least four of the six target intents in 91.7\% of evaluated dialogue states, compared with 22.9\% for the strongest external baseline. These results establish per-turn intent control as a complementary dimension to response fidelity in user simulation.

Problem

• user simulation
• interaction intent
• controllable generation
• dialogue systems
• intent control

Innovation

• controllable user simulation
• interaction intent
• intent-conditioned generation
• directive alignment
• reinforcement learning with intent calibration