Newest papers

LatComm: Reusable Context Compression for Latent Communication between Large Language Models

Huawei Lin, Jinting Liu, Tan Yu, Weijie Zhao · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26840 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Sistemas que combinam vários grandes modelos de linguagem (LLMs) muitas vezes usam um contexto para muitas perguntas. Quando o contexto é transmitido entre modelos como texto, o modelo receptor processa todos os seus tokens novamente para cada pergunta. Evitar esse custo requer uma representação de contexto que é compacta, reutilizável em todas as perguntas e legível por um LLM diferente. Os métodos de compressão e comunicação latente existentes fornecem apenas algumas dessas propriedades. Introduzimos o LatComm, uma interface de comunicação latente que fornece todos os três separando a codificação da leitura: o contexto é codificado uma vez sem a pergunta, e o receptor seleciona o que ler para cada pergunta. Um codificador de contexto congelado LLM comprime o contexto em um pacote de tamanho fixo e o envia para

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

Abstract

Systems that combine several large language models (LLMs) often use one context for many questions. When the context is passed between models as text, the receiving model processes all of its tokens again for every question. Avoiding this cost requires a context representation that is compact, reusable across questions, and readable by a different LLM. Existing compression and latent communication methods each provide only some of these properties. We introduce LatComm, a Latent Communication interface that provides all three by separating encoding from reading: the context is encoded once without the question, and the receiver selects what to read for each question. A frozen context encoder LLM compresses the context into a fixed-size packet and sends it to a different frozen receiver LLM, which reads the packet through cross-attention, with gates computed from its prompt. On 12 multiple-choice question-answering datasets, LatComm improves average accuracy over single-vector baselines by up to 10.2 points. On QuALITY, it reduces the receiver's input by 57x and answers the questions of an article 3x faster.

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

Topics: Computer science, Context (archaeology), Encoding (memory), Context model, Network packet, Language model

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