Asymmetric Looped Transformers: Fixed Global Representations for an Iterative Decoder

Cheng Liang, Tengfei Zhang, Wenjie DU, Jianyu Wu, Changlian Ma, Shuchen Pu, Ya Zhang, Yanfeng Wang · 2026 · arXiv

DOI: 10.48550/arXiv.2610.28489 · open-access PDF

Hindi

लपेटे ट्रांसफार्मर साझा परतों को बार-बार लागू करके प्रभावी गहराई को बढ़ाते हैं, जो पैरामीटर की संख्या में वृद्धि किए बिना गणना को स्केलिंग करने का एक तरीका प्रदान करते हैं. हालांकि, पूर्ण स्टैक लपेटें संदर्भ निर्माण और पुनरावृत्ति राज्य परिष्करण को एक ही पुनरावृत्ति गणना के भीतर रखती हैं, बिना उनके भूमिकाओं को स्पष्ट रूप से अलग करने के।

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

Abstract

Looped Transformers increase effective depth by repeatedly applying shared layers, offering a way to scale computation without increasing parameter count. However, full-stack looping places context construction and iterative state refinement within the same recurrent computation, without explicitly distinguishing their roles. We introduce the Asymmetric Looped Transformer (AsymLoop), which separates single-pass context encoding from recurrent state refinement. A single-pass causal encoder produces layer-specific Global key-value (KV) caches that remain fixed across iterations, while a weight-t

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

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28489, 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.txt

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