PACE

KO | EN

gitlite — search

#typescript #llm #golang #go #deep-learning #docker #macos #java #windows #cli #kubernetes #vue #ios #mcp

(★ 12)

Persona Adaptation through Conversational Elicitation in Human-Robot Interaction

PACE Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# PACE

**Persona Adaptation through Conversational Elicitation in Human-Robot Interaction**

PACE is a framework for dynamically generating and deploying personalized robot personas through short conversational elicitation. Instead of relying on a fixed, developer-written system prompt, PACE allows a humanoid robot to ask open-ended questions, infer psychologically grounded persona attributes, compile them into a structured `PersonaSpec`, and activate the resulting persona in embodied interaction.

This repository accompanies the paper:

> **PACE: Persona Adaptation through Conversational Elicitation in Human-Robot Interaction**

Project page / video demo: https://anonymous.4open.science/w/PACE-CF28/

---

## Overview

PACE consists of three main stages:

1. **Interactive Q&A Persona Elicitation**  
   The robot asks a small set of open-ended anchor questions and generates adaptive follow-up questions when more detail is needed.

2. **Persona Specification Generation**  
   The elicitation transcript is converted into a structured `PersonaSpec` containing psychological and behavioral dimensions such as traits, values, motivation, regulatory orientation, identity claims, and situational policies.

3. **Dynamic Persona Activation**  
   The structured persona is compiled into a system prompt and deployed on the robot. The robot’s verbal responses and facial expressions are conditioned on the generated persona.

---

## Key Features

- Short conversational persona elicitation instead of long psychometric surveys
- Structured persona representation through `PersonaSpec`
- Multi-perspective analysis over traits, values, motivation, orientation, identity, and policies
- Dynamic persona prompt compilation
- Embodied deployment on the Ameca humanoid robot
- Speech output through Amazon Polly
- Persona specification generation using GPT-5.5-mini
- Facial expression selection from seven basic emotion categories
- Context-aware emotion inference for synchronized speech and facial animation

---

## System Architecture

PACE follows an end-to-end pipeline that converts natural user speech into a dynamically activated embodied robot persona.

The pipeline consists of three main stages:

1. **Interactive Q&A**: the humanoid robot elicits persona-relevant traits through open-ended anchor questions and adaptive follow-ups.
2. **Persona Specification Generation**: user responses are transcribed and converted into a structured `PersonaSpec` containing traits, values, motivation, orientation, identity, and situational policies.
3. **Dynamic Persona Activation**: the generated `PersonaSpec` is compiled into a persona prompt, injected into the LLM agent state, and expressed through robot speech and synchronized facial animation.

PACE

PACE Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

PACE

PACE Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README