DENIAHL: In-Context Features Influence LLM Needle-In-A-Haystack Abilities

TL;DR

This paper introduces DENIAHL, a synthetic benchmark to analyze factors affecting language models' ability to recall specific information from long contexts, revealing how features like data type and size influence performance.

Contribution

The study develops a systematic benchmark, DENIAHL, to evaluate how various features beyond context length impact LLMs' NIAH abilities, expanding previous research.

Findings

GPT-3.5 outperforms LLaMA 2-7B on DENIAHL

Recall performance drops with increased item size

Changing data type from numbers to letters affects recall

Abstract

The Needle-in-a-haystack (NIAH) test is a general task used to assess language models' (LMs') abilities to recall particular information from long input context. This framework however does not provide a means of analyzing what factors, beyond context length, contribute to LMs' abilities or inabilities to separate and recall needles from their haystacks. To provide a systematic means of assessing what features contribute to LMs' NIAH capabilities, we developed a synthetic benchmark called DENIAHL (Data-oriented Evaluation of NIAH for LLM's). Our work expands on previous NIAH studies by ablating NIAH features beyond typical context length including data type, size, and patterns. We find stark differences between GPT-3.5 and LLaMA 2-7B's performance on DENIAHL, and drops in recall performance when features like item size are increased, and to some degree when data type is changed from numbers to letters. This has implications for increasingly large context models, demonstrating factors beyond item-number impact NIAH capabilities.