Attention Basin: Why Contextual Position Matters in Large Language Models
This addresses a critical issue for LLM users by improving performance in retrieval-augmented and in-context learning scenarios, though it is incremental as it builds on known positional biases.
The paper tackles the problem of positional bias in Large Language Models (LLMs), where performance is sensitive to the contextual position of information, and introduces Attention-Driven Reranking (AttnRank) to reorder inputs based on attention preferences, achieving substantial improvements across 10 models on tasks like multi-hop QA and few-shot learning.
The performance of Large Language Models (LLMs) is significantly sensitive to the contextual position of information in the input. To investigate the mechanism behind this positional bias, our extensive experiments reveal a consistent phenomenon we term the attention basin: when presented with a sequence of structured items (e.g., retrieved documents or few-shot examples), models systematically assign higher attention to the items at the beginning and end of the sequence, while neglecting those in the middle. Crucially, our analysis further reveals that allocating higher attention to critical information is key to enhancing model performance. Based on these insights, we introduce Attention-Driven Reranking (AttnRank), a two-stage framework that (i) estimates a model's intrinsic positional attention preferences using a small calibration set, and (ii) reorders retrieved documents or few-shot examples to align the most salient content with these high-attention positions. AttnRank is a model-agnostic, training-free, and plug-and-play method with minimal computational overhead. Experiments on multi-hop QA and few-shot in-context learning tasks demonstrate that AttnRank achieves substantial improvements across 10 large language models of varying architectures and scales, without modifying model parameters or training procedures.