Boosting Long-Context Management via Query-Guided Activation Refilling

TL;DR

This paper introduces ACRE, a query-guided activation refilling method that enhances long-context processing in language models by dynamically balancing global and local information, leading to better performance and efficiency.

Contribution

ACRE constructs a bi-layer KV cache with proxying between global and local caches, enabling dynamic refilling based on query needs, which improves long-context understanding.

Findings

Improves accuracy on long-context datasets

Reduces computational overhead compared to existing methods

Effectively balances global and local information retrieval

Abstract

Processing long contexts poses a significant challenge for large language models (LLMs) due to their inherent context-window limitations and the computational burden of extensive key-value (KV) activations, which severely impact efficiency. For information-seeking tasks, full context perception is often unnecessary, as a query's information needs can dynamically range from localized details to a global perspective, depending on its complexity. However, existing methods struggle to adapt effectively to these dynamic information needs. In the paper, we propose a method for processing long-context information-seeking tasks via query-guided Activation Refilling (ACRE). ACRE constructs a Bi-layer KV Cache for long contexts, where the layer-1 (L1) cache compactly captures global information, and the layer-2 (L2) cache provides detailed and localized information. ACRE establishes a proxying relationship between the two caches, allowing the input query to attend to the L1 cache and dynamically refill it with relevant entries from the L2 cache. This mechanism integrates global understanding with query-specific local details, thus improving answer decoding. Experiments on a variety of long-context information-seeking datasets demonstrate ACRE's effectiveness, achieving improvements in both performance and efficiency.