Where Relevance Emerges: A Layer-Wise Study of Internal Attention for Zero-Shot Re-Ranking

TL;DR

This paper investigates internal attention signals within transformer layers for zero-shot re-ranking, revealing a universal relevance distribution and proposing a Selective-ICR method that improves efficiency and effectiveness.

Contribution

It provides a comprehensive layer-wise analysis of internal attention, introduces a universal relevance distribution, and proposes a Selective-ICR strategy that enhances re-ranking efficiency without sacrificing performance.

Findings

A universal bell-curve distribution of relevance signals across transformer layers.

Selective-ICR reduces inference latency by 30%-50%.

A zero-shot 8B model matches or outperforms larger models and state-of-the-art methods.

Abstract

Zero-shot document re-ranking with Large Language Models (LLMs) has evolved from Pointwise methods to Listwise and Setwise approaches that optimize computational efficiency. Despite their success, these methods predominantly rely on generative scoring or output logits, which face bottlenecks in inference latency and result consistency. In-Context Re-ranking (ICR) has recently been proposed as an O(1) alternative method. ICR extracts internal attention signals directly, avoiding the overhead of text generation. However, existing ICR methods simply aggregate signals across all layers; layer-wise contributions and their consistency across architectures have been left unexplored. Furthermore, no unified study has compared internal attention with traditional generative and likelihood-based mechanisms across diverse ranking frameworks under consistent conditions. In this paper, we conduct an orthogonal evaluation of generation, likelihood, and internal attention mechanisms across multiple ranking frameworks. We further identify a universal "bell-curve" distribution of relevance signals across transformer layers, which motivates the proposed Selective-ICR strategy that reduces inference latency by 30%-50% without compromising effectiveness. Finally, evaluation on the reasoning-intensive BRIGHT benchmark shows that precisely capturing high-quality in-context attention signals fundamentally reduces the need for model scaling and reinforcement learning: a zero-shot 8B model matches the performance of 14B reinforcement-learned re-rankers, while even a 0.6B model outperforms state-of-the-art generation-based approaches. These findings redefine the efficiency-effectiveness frontier for LLM-based re-ranking and highlight the latent potential of internal signals for complex reasoning ranking tasks. Our code and results are publicly available at https://github.com/ielab/Selective-ICR.