GenArena: How Can We Achieve Human-Aligned Evaluation for Visual Generation Tasks?

2026-02-05

AI Summary

This work addresses the limitations of prevailing evaluation methods for vision generative models, which predominantly rely on pointwise scoring and suffer from stochastic inconsistency and poor alignment with human perception. To overcome these issues, the authors propose GenArena, a unified automated evaluation framework based on pairwise comparisons. GenArena systematically exposes the deficiencies of pointwise scoring and leverages vision-language models as proxy judges to enable efficient and stable model comparisons. The approach significantly improves alignment with human judgments, achieving a Spearman correlation coefficient of 0.86 with human assessments on the authoritative LMArena leaderboard—substantially outperforming pointwise methods (0.36) by over 20%. Furthermore, the framework demonstrates that open-source models can surpass leading closed-source counterparts under this more reliable evaluation paradigm.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

The rapid advancement of visual generation models has outpaced traditional evaluation approaches, necessitating the adoption of Vision-Language Models as surrogate judges. In this work, we systematically investigate the reliability of the prevailing absolute pointwise scoring standard, across a wide spectrum of visual generation tasks. Our analysis reveals that this paradigm is limited due to stochastic inconsistency and poor alignment with human perception. To resolve these limitations, we introduce GenArena, a unified evaluation framework that leverages a pairwise comparison paradigm to ensure stable and human-aligned evaluation. Crucially, our experiments uncover a transformative finding that simply adopting this pairwise protocol enables off-the-shelf open-source models to outperform top-tier proprietary models. Notably, our method boosts evaluation accuracy by over 20% and achieves a Spearman correlation of 0.86 with the authoritative LMArena leaderboard, drastically surpassing the 0.36 correlation of pointwise methods. Based on GenArena, we benchmark state-of-the-art visual generation models across diverse tasks, providing the community with a rigorous and automated evaluation standard for visual generation.

Problem

• visual generation evaluation
• human alignment
• pointwise scoring
• evaluation reliability
• Vision-Language Models

Innovation

• pairwise comparison
• human-aligned evaluation
• visual generation
• Vision-Language Models
• GenArena