Lost in the Noise : How Reasoning Models Fail with Contextual Distractors

Seongyun Lee, Yongrae Jo, Minju Seo, Moontae Lee, Minjoon Seo · 2026

Read Lost in the Noise : How Reasoning Models Fail with Contextual Distractors by Seongyun Lee, Yongrae Jo, Minju Seo, Moontae Lee, Minjoon Seo (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-07226.txt.

Description

Seongyun Lee and co-authors posted a 2026 arXiv preprint. The preprint deals with lost in the noise. It came out on arXiv in 2026 as a PAPERS preprint. Recent advances in reasoning models and agentic AI systems have led to an increased reliance on diverse external information. However, this shift introduces input contexts that are inherently noisy, a reality that current sanitized benchmarks fail to capture.

The authors introduce NoisyBench , a comprehensive benchmark that systematically evaluates model robustness across 11 datasets in RAG, reasoning. Alignment, and tool-use tasks against diverse noise types, including random documents, irrelevant chat histories, and hard negative distractors. Readers in the field get a short, focused account of the problem and the steps the authors take.

Catalog record

CREATORSeongyun Lee, Yongrae Jo, Minju Seo, Moontae Lee, Minjoon Seo
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE114 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-07226

Browse the full library · About the Tower of Babel