October 5, 2025Open Access

Have we unified image generation and understanding yet? An empirical study of GPT-4o's image generation ability

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Empirical analysis shows GPT-4o's limitations in achieving semantic synthesis and instruction adherence.

GPT-4o fails to consistently apply domain knowledge during image generation and editing tasks.
Evaluation metrics indicate persistent limitations, especially in conditional reasoning and instruction fidelity.
The systematic study across three dimensions reveals significant gaps in GPT-4o's multimodal generation capabilities.
Findings suggest a need for enhanced benchmarks and training strategies to improve context-aware generation.

Li et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

October 5, 2025Open Access

View Full Paper

Ask AI

Bookmark

Authors

Empirical analysis shows GPT-4o's limitations in achieving semantic synthesis and instruction adherence.

GPT-4o fails to consistently apply domain knowledge during image generation and editing tasks.
Evaluation metrics indicate persistent limitations, especially in conditional reasoning and instruction fidelity.
The systematic study across three dimensions reveals significant gaps in GPT-4o's multimodal generation capabilities.
Findings suggest a need for enhanced benchmarks and training strategies to improve context-aware generation.

Li et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark