Minkyun Seo, Hyungtae Lim, Jaesik Park · 2026 · arXiv
DOI: 10.48550/arXiv.2610.31455 · open-access PDF
Transformator awan titik bertujuan untuk menangkap geometri lokal yang terperinci dan hubungan di seluruh adegan dengan biaya komputasi yang dapat dikelola. Dua pendekatan umum adalah perhatian jendela dan perhatian global. Mereka berbeda dalam cara mereka memungkinkan interaksi di seluruh adegan, yang mengacu pada pertukaran informasi antara token di seluruh adegan. Perhatian jendela mengurangi biaya komputasi dengan membatasi perhatian diri ke kelompok token yang disebut jendela; dalam desain seri, setiap jendela berisi sejumlah token berturut-turut dari urutan yang diatur secara spasial. Karena setiap blok hanya langsung menghubungkan token di dalamnya.
Machine translation (Meta M2M100); the English original follows.
Point cloud transformers aim to capture detailed local geometry and relationships across a scene at a manageable computational cost. Two common approaches are window attention and global attention. They differ in how they enable scene-wide interaction, which refers to information exchange among tokens across a scene. Window attention reduces computational cost by restricting self-attention to groups of tokens called windows; in serialized designs, each window contains a fixed number of consecutive tokens from a spatially ordered sequence. Because each block only directly connects tokens within
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=2610.31455, 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