Interact2Ar: Full-Body Human-Human Interaction Generation via Autoregressive Diffusion Models

2025-12-22

AI Summary

Existing methods struggle to jointly model full-body motion, fine-grained hand articulation, and dynamic inter-participant coordination in social interactions, while diffusion models suffer from limited reactivity and temporal controllability. This paper proposes the first multi-branch autoregressive diffusion model tailored for full-body interpersonal interaction. It introduces a parallel hand-motion branch to enhance hand fidelity, incorporates a memory-augmented large-context-window mechanism enabling real-time perturbation response, long-horizon temporal modeling, and natural extension to multi-person scenarios, and establishes a dedicated evaluation framework for full-body social interaction. Experiments demonstrate state-of-the-art performance in both quantitative and qualitative assessments—significantly improving hand-detail realism and interaction coordination. The approach establishes a new paradigm for text-driven, high-fidelity interpersonal motion generation.

Technology Category

• Humans and AI: Interaction Techniques and Devices
• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Assisted, interactive, and conversational search

Abstract

Generating realistic human-human interactions is a challenging task that requires not only high-quality individual body and hand motions, but also coherent coordination among all interactants. Due to limitations in available data and increased learning complexity, previous methods tend to ignore hand motions, limiting the realism and expressivity of the interactions. Additionally, current diffusion-based approaches generate entire motion sequences simultaneously, limiting their ability to capture the reactive and adaptive nature of human interactions. To address these limitations, we introduce Interact2Ar, the first end-to-end text-conditioned autoregressive diffusion model for generating full-body, human-human interactions. Interact2Ar incorporates detailed hand kinematics through dedicated parallel branches, enabling high-fidelity full-body generation. Furthermore, we introduce an autoregressive pipeline coupled with a novel memory technique that facilitates adaptation to the inherent variability of human interactions using efficient large context windows. The adaptability of our model enables a series of downstream applications, including temporal motion composition, real-time adaptation to disturbances, and extension beyond dyadic to multi-person scenarios. To validate the generated motions, we introduce a set of robust evaluators and extended metrics designed specifically for assessing full-body interactions. Through quantitative and qualitative experiments, we demonstrate the state-of-the-art performance of Interact2Ar.

Problem

• Generates realistic full-body human-human interactions with detailed hand motions.
• Uses autoregressive diffusion to capture reactive and adaptive interaction dynamics.
• Enables applications like temporal composition and real-time adaptation to disturbances.

Innovation

• Autoregressive diffusion model for full-body human interactions
• Parallel branches for detailed hand kinematics generation
• Memory technique for adaptive large context windows