ICG: Improving Cover Image Generation via MLLM-based Prompting and Personalized Preference Alignment
Quick Answer
The ICG framework enhances personalized cover image generation by integrating MLLM-based prompting with user preference alignment, significantly improving image quality and user engagement metrics without requiring labeled data.
Quick Take
The ICG framework enhances personalized cover image generation by integrating MLLM-based prompting with user preference alignment, significantly improving image quality and user engagement metrics without requiring labeled data. It employs a multi-reward learning strategy to optimize performance across various downstream tasks.
Key Points
- ICG utilizes meta tokens to extract semantic features from titles and reference images.
- The framework employs a multi-reward learning strategy combining aesthetic and personalized rewards.
- ICG bridges MLLMs and diffusion models for seamless end-to-end training.
- Experiments show improved image quality and personalization, enhancing user appeal.
- ICG is compatible with common checkpoints and requires no ground-truth labels.
Paper Resources
Article Content
From source RSS / original summaryarXiv:2605. 27374v1 Announce Type: new Abstract: Recent advances in multimodal large language models (MLLMs) and diffusion models (DMs) have opened new possibilities for AI-generated content. Yet, personalized cover image generation remains underexplored, despite its critical role in boosting user engagement on digital platforms. We propose ICG, a novel framework that integrates MLLM-based prompting with personalized preference alignment to generate high-quality, contextually relevant covers.
ICG extracts semantic features from item titles and reference images via meta tokens, refines them with user embeddings, and injects the resulting personalized context into the diffusion model. To address the lack of labeled supervision, we adopt a multi-reward learning strategy that combines public aesthetic and relevance rewards with a personalized preference model trained from user behavior.
Unlike prior pipelines relying on handcrafted prompts and disjointed modules, ICG employs an adapter to bridge MLLMs and diffusion models for end-to-end training. Experiments demonstrate that ICG significantly improves image quality, semantic fidelity, and personalization, leading to stronger user appeal and offline recommendation accuracy in downstream tasks.
As a plug-and-play adapter bridging MLLMs and diffusion models, ICG is compatible with common checkpoints and requires no ground-truth labels during optimization.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
The REFLECT benchmark reveals that current LLM judges are unreliable, achieving below 55% accuracy in evaluating reasoning and evidence use, highlighting the need for improved evaluation methods for deep research agents.