TactiDex: A Real-World Tactile-Guided Benchmark for Human-Like Dexterous Manipulation

2026-07-10

AI Summary

This work addresses the limitations of existing dexterous manipulation approaches, which predominantly rely on kinematic imitation and lack realistic tactile feedback, thereby struggling to achieve human-like contact-level control. To bridge this gap, the authors introduce TactiDex—the first real-world, tactile-guided benchmark for dexterous manipulation—featuring synchronized recordings of full-hand tactile signals, multi-granularity kinematics, and object states. Built upon this benchmark, they propose TactiSkill, a tactile-driven transfer framework that innovatively leverages structured tactile signals as supervision to design a three-component tactile reward mechanism. This mechanism unifies task objectives, human behavioral priors, and physical contact constraints. Experiments demonstrate that the proposed approach significantly improves both success rates and physical plausibility in single- and dual-hand manipulation tasks, establishing a foundation for tactile-perception-driven dexterous manipulation.

Abstract

Tactile feedback is fundamental to Hand-Object Interaction (HOI), governing contact formation, force regulation, and stable manipulation, making it essential for achieving true human-like dexterous manipulation. Yet, current human-to-robot dexterous transfer pipelines primarily rely on kinematic trajectories, resulting in motion imitation without physically grounded interaction. To address this, we introduce TactiDex, a real-world tactile-guided benchmark specifically designed to move dexterous manipulation beyond kinematic mimicry toward contact-level human-likeness. TactiDex provides a comprehensive dataset that elegantly aligns whole-hand tactile signals with multi-granularity kinematic and object states, coupled with standardized evaluation metrics. Building upon this data paradigm, we propose a tactile-driven transfer framework that effectively translates human demonstrations into physically plausible robotic execution. We introduce TactiSkill, a framework built upon a novel tri-component tactile reward that innovatively uses tactile signals as structured supervision. This reward unifies guidance, human-like alignment, and contact constraints into a single objective. Through comprehensive experiments on both single and bimanual tasks, we demonstrate that TactiSkill achieves superior performance in manipulation success and physical realism. This work lays a crucial foundation for advancing tactile-aware dexterous manipulation. Our project page at https://tactidex.github.io/.

Problem

• tactile feedback
• dexterous manipulation
• human-to-robot transfer
• hand-object interaction
• contact-level realism

Innovation

• tactile-guided manipulation
• dexterous manipulation
• human-to-robot transfer
• tactile reward
• contact-level realism