arXiv cs.AI· Zhengfeng Zheng, Minghong Fang·· 12 小时前AI 评分31
EmbGA：在嵌入空间用遗传算法恢复文生图模型提示词中隐藏的概念
Beyond the Prompt: Recovering Hidden Concepts in Text-to-Image Models
AI 导读
论文提出 EmbGA，一种在 token 嵌入空间运行的遗传算法，用于恢复文生图模型生成图像中存在但提示词未提及的概念。其交叉在父代嵌入间插值、变异将嵌入移向随机词表嵌入，每个子代先投影回离散 token 再评分，保证候选均为合法 token 序列。在两个基准、两种探索预算下对比六种优化器，EmbGA 在大预算时对齐度最高、在自然图像上恢复出最多有依据的 token，且比语言模型引导的搜索快得多。
正文 · 原文
Abstract:Text-to-image models can generate visual concepts that their prompts never mention, and recent work shows that such associations can be implanted so that ordinary prompts trigger them. Auditing this behavior requires recovering the description that a prompt omits. Existing prompt optimizers are not well suited to this task. Gradient-based methods require backward access to the scoring encoder. Discrete evolutionary search treats semantically unrelated tokens as equally close. Language-model-guided search requires one inference call per candidate, which limits exploration. This paper presents EmbGA, an embedding-space genetic algorithm that recovers concepts present in a set of images but absent from the prompt that generated them. EmbGA keeps the population structure of genetic search but applies its operators in the token embedding space. Crossover interpolates between parent embeddings, and mutation moves an embedding toward a randomly drawn vocabulary embedding. Each offspring is then projected back to discrete tokens before scoring, so every evaluated candidate is a valid token sequence. We evaluate EmbGA on two benchmarks against six optimizers under two exploration budgets. EmbGA achieves the highest alignment when the exploration budget is large, recovers the most grounded tokens on natural images, and runs much faster than language-model-guided search.
| Subjects: | Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.27473 [cs.CR] |
| (or arXiv:2610.27473v1 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2610.27473 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Minghong Fang [view email]
[v1]
Tue, 20 Oct 2026 01:42:42 UTC (2,543 KB)
来源：arXiv cs.AI · arxiv.org