🛰️ Daily AI Frontier
‹ back to 2026-09-03

Language Models Can Control Their Own Attention

Research Efficiency & Systems

Ranking

Overall 87
Content 95
Popularity 68

Observed public metrics from 1 member.

Merged summary

TL;DR - Declarative Attention lets language models explicitly indicate which context regions they need during generation, allowing the inference engine to avoid reading most of the KV cache. On long-context tasks, it substantially reduced attended tokens with modest accuracy losses.

  • The protocol uses three model-declared modes: <global> for full context, <focus> for a specific region, and <local> for recent output only.
  • Unlike proxy-based token selection, the model specifies relevant context intrinsically through chain-of-thought declarations parsed like tool calls.
  • Across 15 zero-shot long-context tasks, attended tokens fell by 52.0% for Gemma-4-31B and 31.1% for Qwen-3.6-27B.
  • Accuracy declined by 1.27 and 2.75 percentage points, respectively, with losses reported to shrink as model scale increases.

Sources (1)

Language Models Can Control Their Own Attention

arXiv cs.CL Namgyu Ho, Huzama Ahmad, Woosung Koh, Se-Young Yun, Tal Schuster, Cicero Nogueira dos Santos 2026-09-02 arXiv:2609.02737
Public signals Hugging Face upvotes 78
Providers: Hugging Face · Upvotes 78 OpenAlex · N/A Publisher · N/A Semantic Scholar · N/A X · N/A Fetched 2026-09-25 14:24:33.223286 UTC

TL;DR - Declarative Attention lets language models explicitly indicate which context regions they need during generation, allowing the inference engine to avoid reading most of the KV cache. On long-context tasks, it substantially reduced attended tokens with modest accuracy losses.

  • The protocol uses three model-declared modes: <global> for full context, <focus> for a specific region, and <local> for recent output only.
  • Unlike proxy-based token selection, the model specifies relevant context intrinsically through chain-of-thought declarations parsed like tool calls.
  • Across 15 zero-shot long-context tasks, attended tokens fell by 52.0% for Gemma-4-31B and 31.1% for Qwen-3.6-27B.
  • Accuracy declined by 1.27 and 2.75 percentage points, respectively, with losses reported to shrink as model scale increases.
item →