Positional Biases Shift as Inputs Approach Context Window Limits

Blerta Veseli, Julian Chibane, Mariya Toneva, Alexander Koller

Keywords: Long-context understanding, positional biases

TL;DR: This paper examines how input length, relative to a model’s context window, affects positional biases in LLMs.

Abstract:

Large Language Models (LLMs) often struggle to use information across long inputs effectively.
Prior work has identified positional biases, such as the Lost in the Middle (LiM) effect, where models perform better when information appears at the beginning (primacy bias) or end (recency bias) of the input, rather than in the middle.
However, long-context studies have not consistently replicated these effects, raising questions about their intensity and the conditions under which they manifest.
To address this, we conducted a comprehensive analysis using relative rather than absolute input lengths, defined with respect to each model’s context window.
Our findings reveal that the LiM effect is strongest when inputs occupy up to 50\% of a model’s context window.
Beyond that, the primacy bias weakens, while recency bias remains relatively stable.
This effectively eliminates the LiM effect; instead, we observe a distance-based bias, where model performance is better when relevant information is closer to the end of the input.
Furthermore, our results suggest that successful retrieval is a prerequisite for reasoning in LLMs, and that the observed positional biases in reasoning are largely inherited from retrieval. These insights have implications for long-context tasks, the design of future LLM benchmarks, and evaluation methodologies for LLMs handling extended inputs.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the COLM Code of Ethics on https://colmweb.org/CoE.html

Author Guide: I certify that this submission complies with the submission instructions as described on https://colmweb.org/AuthorGuide.html

Submission Number: 1047