Matteo Cercola, Katie Z Luo, Milan Ganai, John Alora, Simone Formentin, Marco Pavone · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28166 · open-access PDF
Kebijakan Vision-Language-Action (VLA) semakin mempertimbangkan alasan sebelum bertindak, tetapi model argumen yang ada menyemprotkan keputusan pertengahan ini menjadi satu representasi. Ketika beberapa rencana berlaku, komitmen ini menyembunyikan alternatif yang bisa dipilih atau diperbaiki. Kami memperkenalkan Generative Plan Latent (GPL), yang mewakili keadaan argumen sebagai distribusi kondisional rendah-dimensi di atas rencana, seperti objek, pemahaman, atau rute yang akan digunakan. Ini membuat keadaan argumen dapat ditangani: tanpa melatih ulang VLA, rencana dapat dicampur (coverage), dipilih atas permintaan (steerin)
Machine translation (Meta M2M100); the English original follows.
Vision-Language-Action (VLA) policies increasingly reason before acting, but existing reasoning models compress this intermediate decision into a single representation. When several plans are valid, this commitment hides alternatives that could otherwise be selected or revised. We introduce the Generative Plan Latent (GPL), which represents the reasoning state as a low-dimensional conditional distribution over plans, such as which object, grasp, or route to use. This makes the reasoning state addressable: without retraining the VLA, plans can be sampled (coverage), selected on request (steerin
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28166, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange