Correcting Visual Blur Induced by Attention Distraction to Reduce Hallucinations: Algorithm and Theory

Quanjiang Li, Zhiming Liu, Wei Luo, Tingjin Luo, Chenping Hou · 2026

Read Correcting Visual Blur Induced by Attention Distraction to Reduce Hallucinations: Algorithm and Theory by Quanjiang Li, Zhiming Liu, Wei Luo, Tingjin Luo, Chenping Hou (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-2605-24602.txt.

Description

Quanjiang Li and others trace object hallucinations in multimodal models to a kind of focus failure. The pattern recalls a person describing a scene while distracted. In models the trouble shows as uneven focus across heads and fading focus on image tokens as decoding goes on. Theory links scattered focus to weaker generalization. Their training-free fix, AFIP, enriches focus across heads and strengthens visual grounding with past focus. It cuts hallucinations across several benchmarks and models, with no extra training.

Catalog record

CREATORQuanjiang Li, Zhiming Liu, Wei Luo, Tingjin Luo, Chenping Hou
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE98419
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.24602

Browse the full library · About the Tower of Babel