LatComm: Reusable Context Compression for Latent Communication between Large Language Models
arXiv cs.AI · 20 Oct 2026, 03:58 AEST
LatComm introduces a new interface for efficient communication between large language models. It allows for compact, reusable context representation that enhances performance on multiple-choice question-answering tasks.
Key facts
- Systems that combine several large language models (LLMs) often use one context for many questions.
- 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.
- On 12 multiple-choice question-answering datasets, LatComm improves average accuracy over single-vector baselines by up to 10.2 points.
Read aloud
LatComm introduces a new interface for efficient communication between large language models. It allows for compact, reusable context representation that enhances performance on multiple-choice question-answering tasks. Systems that combine several large language models L L M s often use one context for many questions. 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. On twelve multiple-choice question-answering datasets, LatComm improves average accuracy over single-vector baselines by up to ten point two points.
Sources in this cluster
- LatComm: Reusable Context Compression for Latent Communication between Large Language Models
arXiv cs.AI · 20 Oct 2026, 03:58 AEST