Newest papers

Beyond the Prompt: Recovering Hidden Concepts in Text-to-Image Models

Zhengfeng Zheng, Minghong Fang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.27473 · cited by 0 works (OpenAlex) · open-access PDF

Korean

텍스트-to-Image 모델은 그들의 프롬프트가 언급하지 않는 시각적 개념을 생성 할 수 있으며 최근의 연구는 그러한 연관성이 일반 프롬프트가 그들을 촉발시킬 수 있음을 보여줍니다. 이 행동을 감사하는 것은 프롬프트가 놓치고있는 설명을 복구해야합니다. 기존의 프롬프트 최적화기는이 작업에 적합하지 않습니다. Gradient 기반의 방법은 점수 인코더에 뒤로 액세스 할 수 있습니다. 차별화 된 진화적 검색은 정의적으로 관련이없는 토큰을 동등하게 가까이 다루고 있습니다. 언어 모델 지시 검색은 후보 당 하나의 추론 호출을 필요로합니다.이 논문은 EmbGA를 제시합니다.이 논문은 6 개의 이미지 세트에 존재하는 개념을 복구

Machine translation (Meta M2M100); the English original follows.

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.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Set (abstract data type), Security token, Crossover, Population, Embedding

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange