MANTA -- Model Adapter Native generations that's Affordable

TL;DR

MANTA introduces a flexible, cost-effective model adapter approach that enhances image task diversity and quality, outperforming existing systems while maintaining affordability and practical hardware considerations.

Contribution

The paper presents MANTA, a novel model adapter composition method addressing hardware and cost constraints, improving image task diversity and quality.

Findings

Achieves 94% win rate in task diversity

Achieves 80% win rate in task quality

Demonstrates potential in synthetic data and creative art domains

Abstract

The presiding model generation algorithms rely on simple, inflexible adapter selection to provide personalized results. We propose the model-adapter composition problem as a generalized problem to past work factoring in practical hardware and affordability constraints, and introduce MANTA as a new approach to the problem. Experiments on COCO 2014 validation show MANTA to be superior in image task diversity and quality at the cost of a modest drop in alignment. Our system achieves a $94\%$ win rate in task diversity and a $80\%$ task quality win rate versus the best known system, and demonstrates strong potential for direct use in synthetic data generation and the creative art domains.