AnyBottle method creates compact, task-specific AI concept models

Researchers have developed AnyBottle, a novel method for creating compact, task-specific concept bottleneck models (CBMs). Unlike previous annotation-free variants that used large, static concept vocabularies, AnyBottle utilizes a frozen backbone and an unsupervised concept pool, such as a sparse autoencoder. A black-box teacher model guides the selection process by identifying concepts that best explain the bottleneck's failures, thereby creating smaller, more inspectable bottlenecks. This approach has demonstrated effectiveness across various vision and text datasets, yielding fewer concepts and higher consistency compared to existing methods. AI

IMPACT Enables more efficient and interpretable AI models by reducing concept vocabulary size.

RANK_REASON The cluster contains a research paper detailing a new method for AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• AnyBottle
• arXiv
• CatalyzeX
• Concept Bottleneck Models
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• ScienceCast
• Sparse Autoencoder

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →