Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2512.20666 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918261799518208 |
|---|---|
| author | Jeong, Hayeon Lee, Jong-Seok |
| author_facet | Jeong, Hayeon Lee, Jong-Seok |
| contents | Text-to-image diffusion models have drawn significant attention for their ability to generate diverse and high-fidelity images. However, when generating from multi-concept prompts, one concept token often dominates the generation, suppressing the others-a phenomenon we term the Dominant-vs-Dominated (DvD) imbalance. To systematically analyze this imbalance, we introduce DominanceBench and examine its causes from both data and architectural perspectives. Through various experiments, we show that the limited instance diversity in training data exacerbates the inter-concept interference. Analysis of cross-attention dynamics further reveals that dominant tokens rapidly saturate attention, progressively suppressing others across diffusion timesteps. In addition, head ablation studies show that the DvD behavior arises from distributed attention mechanisms across multiple heads. Our findings provide key insights into generative collapse, advancing toward more reliable and controllable text-to-image generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_20666 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Dominating vs. Dominated: Generative Collapse in Diffusion Models Jeong, Hayeon Lee, Jong-Seok Machine Learning Artificial Intelligence Text-to-image diffusion models have drawn significant attention for their ability to generate diverse and high-fidelity images. However, when generating from multi-concept prompts, one concept token often dominates the generation, suppressing the others-a phenomenon we term the Dominant-vs-Dominated (DvD) imbalance. To systematically analyze this imbalance, we introduce DominanceBench and examine its causes from both data and architectural perspectives. Through various experiments, we show that the limited instance diversity in training data exacerbates the inter-concept interference. Analysis of cross-attention dynamics further reveals that dominant tokens rapidly saturate attention, progressively suppressing others across diffusion timesteps. In addition, head ablation studies show that the DvD behavior arises from distributed attention mechanisms across multiple heads. Our findings provide key insights into generative collapse, advancing toward more reliable and controllable text-to-image generation. |
| title | Dominating vs. Dominated: Generative Collapse in Diffusion Models |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2512.20666 |