iLoRA: Bayesian Low-Rank Adaptation with Latent Interaction Graphs for Microbiome Diagnosis

Yang Song, Yixuan Zhang, Lingfa Meng, Tongyuan Hu, Haizhou Shi, Hao Wang, Samir Bhatt, Hengguan Huang · 2026

Read iLoRA: Bayesian Low-Rank Adaptation with Latent Interaction Graphs for Microbiome Diagnosis by Yang Song, Yixuan Zhang, Lingfa Meng, Tongyuan Hu, Haizhou Shi, Hao Wang, Samir Bhatt, Hengguan Huang (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-2605-30179.txt.

Description

Yang Song and seven co-authors introduce iLoRA, a Bayesian graph-conditioned low-rank adaptation framework. It exposes latent interactions driving labels, which standard LoRA hides with a static low-rank update. iLoRA builds a latent interaction graph out of the input and makes input-conditioned updates, learning prediction and interaction structure jointly. For microbiome diagnosis, where disease depends on species levels and microbe cross-talk, it beats LoRA and Bayesian baselines. Across question answering and multi-cohort bowel disease diagnosis, it recovers human-aligned graphs with calibrated uncertainty.

Catalog record

CREATORYang Song, Yixuan Zhang, Lingfa Meng, Tongyuan Hu, Haizhou Shi, Hao Wang, Samir Bhatt, Hengguan Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE77988
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.30179

Browse the full library · About the Tower of Babel