DySCO: Dynamic Attention-Scaling Decoding for Long-Context Language Models

TL;DR

DySCO is a training-free decoding algorithm that enhances long-context reasoning in language models by dynamically adjusting attention to task-relevant tokens using retrieval heads, leading to significant performance improvements.

Contribution

It introduces DYSCO, a novel, training-free decoding method that leverages retrieval heads to improve long-context reasoning in existing language models.

Findings

DYSCO improves performance on long-context reasoning benchmarks by up to 25%.

The method is applicable to any off-the-shelf language model.

Dynamic attention rescaling and retrieval-head guided selection are key to its effectiveness.

Abstract

Understanding and reasoning over long contexts is a crucial capability for language models (LMs). Although recent models support increasingly long context windows, their accuracy often deteriorates as input length grows. In practice, models often struggle to keep attention aligned with the most relevant context throughout decoding. In this work, we propose DYSCO, a novel decoding algorithm for improving long-context reasoning. DYSCO leverages retrieval heads--a subset of attention heads specialized for longcontext retrieval--to identify task-relevant tokens at each decoding step and explicitly up-weight them. By doing so, DYSCO dynamically adjusts attention during generation to better utilize relevant context. The method is training-free and can be applied directly to any off-the-shelf LMs. Across multiple instruction-tuned and reasoning models, DYSCO consistently improves performance on challenging long-context reasoning benchmarks, yielding relative gains of up to 25% on MRCR and LongBenchV2 at 128K context length with modest additional compute. Further analysis highlights the importance of both dynamic attention rescaling and retrievalhead guided selection for the effectiveness of the method, while providing interpretability insights into decoding-time attention behavior. Our code is available at https://github.com/princeton-pli/DySCO.