LLM In-Context Recall is Prompt Dependent

TL;DR

This paper investigates how the in-context recall ability of large language models depends on prompt content and biases, using a needle-in-a-haystack method to analyze performance patterns and suggest improvements.

Contribution

It introduces a systematic evaluation of LLMs' in-context recall using a novel needle-in-a-haystack approach, revealing factors affecting recall performance and potential avenues for enhancement.

Findings

Recall performance varies with haystack length and needle placement.

Biases in training data can impair recall accuracy.

Adjustments in architecture and training improve recall capabilities.

Abstract

The proliferation of Large Language Models (LLMs) highlights the critical importance of conducting thorough evaluations to discern their comparative advantages, limitations, and optimal use cases. Particularly important is assessing their capacity to accurately retrieve information included in a given prompt. A model's ability to do this significantly influences how effectively it can utilize contextual details, thus impacting its practical efficacy and dependability in real-world applications. Our research analyzes the in-context recall performance of various LLMs using the needle-in-a-haystack method. In this approach, a factoid (the "needle") is embedded within a block of filler text (the "haystack"), which the model is asked to retrieve. We assess the recall performance of each model across various haystack lengths and with varying needle placements to identify performance patterns. This study demonstrates that an LLM's recall capability is not only contingent upon the prompt's content but also may be compromised by biases in its training data. Conversely, adjustments to model architecture, training strategy, or fine-tuning can improve performance. Our analysis provides insight into LLM behavior, offering direction for the development of more effective applications of LLMs.