aiqbench analysis
RAIGen introduces a novel framework for identifying rare and underrepresented attributes in text-to-image diffusion models without relying on predefined fairness categories. By leveraging Matryoshka Sparse Autoencoders to analyze neuron activations, the method effectively audits model representations beyond standard demographic biases to uncover diverse cultural or stylistic features. This research advances the field of machine learning interpretability by enabling systematic discovery and amplification of minority data distributions in generative AI systems.
RAIGen: Rare Attribute Identification in Text-to-Image Generative Models
arXiv:2602.06806v3 Announce Type: replace-cross Abstract: Text-to-image diffusion models achieve impressive generation quality but inherit and amplify training-data biases, skewing coverage of semantic attributes. Prior work addresses this in two ways. Closed-set approaches mitigate biases in predefined fairness categories (e.g., gender, race), assuming socially salient minority attributes are known a priori. Open-set approaches frame the task as bias identification, highlighting majority attributes that dominate outputs. Both overlook a complementary task: uncovering rare or minority features underrepresented in the data distribution (social, cultural, or stylistic) yet still encoded in model representations. We introduce RAIGen, the first framework, to our knowledge, for label-free rare-attribute discovery in diffusion models, requiring no predefined minority categories. RAIGen leverages Matryoshka Sparse Autoencoders and a novel minority metric combining neuron activation frequency with semantic distinctiveness to identify interpretable neurons whose top-activating images reveal underrepresented attributes. Experiments show RAIGen discovers attributes beyond fixed fairness categories in Stable Diffusion, scales to larger models such as SDXL, supports systematic auditing across architectures, and enables targeted amplification of rare attributes during generation. The project page is available at https://vssilpa.github.io/RAIGen_webpage/ .
Abstract:Text-to-image diffusion models achieve impressive generation quality but inherit and amplify training-data biases, skewing coverage of semantic attributes. Prior work addresses this in two ways. Closed-set approaches mitigate biases in predefined fairness categories (e.g., gender, race), assuming socially salient minority attributes are known a priori. Open-set approaches frame the task as bias identification, highlighting majority attributes that dominate outputs. Both overlook a complementary task: uncovering rare or minority features underrepresented in the data distribution (social, cultural, or stylistic) yet still encoded in model representations. We introduce RAIGen, the first framework, to our knowledge, for label-free rare-attribute discovery in diffusion models, requiring no predefined minority categories. RAIGen leverages Matryoshka Sparse Autoencoders and a novel minority metric combining neuron activation frequency with semantic distinctiveness to identify interpretable neurons whose top-activating images reveal underrepresented attributes. Experiments show RAIGen discovers attributes beyond fixed fairness categories in Stable Diffusion, scales to larger models such as SDXL, supports systematic auditing across architectures, and enables targeted amplification of rare attributes during generation. The project page is available at this https URL .
Accepted at ICML 2026. Webpage and code available at this https URL
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2602.06806 [cs.CV]
(or arXiv:2602.06806v3 [cs.CV] for this version)
Full text under the publisher's open licence · 204 words