Language Models Can Control Their Own Attention Researchers introduced Declarative Attention (DA), a protocol that lets language models declare which parts of their context they need to attend to, reducing total attended tokens by 52.0% on Gemma-4-31B and 31.1% on Qwen-3.6-27B across 15 long-context tasks, with modest accuracy drops of 1.27 and 2.75 percentage points respectively. The method, described in a paper submitted to arXiv on September 2, 2026, partitions generation into global, focus, and local modes, allowing the inference engine to skip most of the KV cache read. Computer Science Computation and Language Submitted on 2 Sep 2026 Title:Language Models Can Control Their Own Attention View PDF /pdf/2609.02737 HTML experimental https://arxiv.org/html/2609.02737v1 Abstract:Language models spend most of their attention on a small fraction of context, yet they read the entire KV cache to find the few tokens that matter. If the user asks about a previous detail in a 1M-token conversation, global attention layers must scan the full context to generate each token of the reply. A prominent approach mitigates this cost by pre-selecting relevant tokens via lightweight proxy scores, but this extrinsic scoring still incurs O N per step. We take an intrinsic approach motivated by the simple question: wouldn't the model already know which parts of the context are relevant? To this end, we introduce Declarative Attention DA , a protocol that elicits the model to declare where it needs to attend within its chain-of-thought, partitioning generation into three modes: