Newest papers

GPT-6-Astra Understands Touch and Outperform Vision--Tactile--Language--Action Models

Ziyong Wang, Yue Lin, Runxuan Huang, Liang Lin, Dongyu Zhang, Guangrun Wang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28457 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Os modelos multimodais de propósito geral corrigidos mostraram promessa como políticas de ação direta de robôs, mas se eles compreendem a física subjacente às interações do mundo real e podem aplicar esse entendimento para orientar a ação permanece inexplorado. Nós investigamos esta questão usando a manipulação rica em contato como um teste representativo, examinando a compreensão táctil como uma instância concreta desta capacidade mais ampla. A manipulação rica em contato requer interpretação de feedback táctil e referências tácticas para avaliação de tarefas e ações corretivas após a obstrução. As políticas tácticas aprendidas existentes adquirem essas capacidades através de pré-treinamento especializado e otimização de feedback por tarefa, com custos substanciais de dados e treinamento. Nós apresentamos o GPT-as-Tactile-Policy como uma instância

Machine translation (Meta M2M100); the English original follows.

Abstract

General-purpose multimodal models have shown promise as direct robot action policies, yet whether they grasp the physics underlying real-world interactions and can apply that understanding to guide action remains underexplored. We investigate this question using contact-rich manipulation as a representative testbed, examining tactile understanding as a concrete instance of this broader capability. Contact-rich manipulation requires interpreting tactile feedback and adjusting actions after obstruction. Existing learned tactile policies acquire these capabilities through specialized pretraining and per-task optimization, with substantial data and training costs. We present GPT-as-Tactile-Policy, which summarizes expert demonstrations into task guidance and tactile references for contact assessment and corrective action. GPT-6-Astra combines this context with multimodal observations and interaction history to generate control commands and adjust actions through contact feedback. Without additional robotics-specific pretraining or per-task policy fine-tuning, our method achieves 93.8\% mean success across five UniVTAC tasks. Overall performance is comparable to state-of-the-art policies, with higher success on selected tasks. On Insert Hole, tactile feedback improves success from 31\% to 99\%. Execution analysis further demonstrates contact-guided probing and correction, supporting manipulation progress through explicit expert experience and online feedback.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Task (project management), GRASP, Action (physics), Computer science, Human–computer interaction, Context (archaeology)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange