Zero-shot Learning Research Articles

Retrieval-Augmented Primitive Representations for Compositional Zero-Shot Learning

Compositional zero-shot learning (CZSL) aims to recognize unseen attribute-object compositions by learning from seen compositions. Composing the learned knowledge of seen primitives, i.e., attributes or objects, into novel compositions is critical for CZSL. In this work, we propose to explicitly retrieve knowledge of seen primitives for compositional zero-shot learning. We present a retrieval-augmented method, which augments standard multi-path classification methods with two retrieval modules. Specifically, we construct two databases storing the attribute and object representations of training images, respectively. For an input training/testing image, we use two retrieval modules to retrieve representations of training images with the same attribute and object, respectively. The primitive representations of the input image are augmented by using the retrieved representations, for composition recognition. By referencing semantically similar images, the proposed method is capable of recalling knowledge of seen primitives for compositional generalization. Experiments on three widely-used datasets show the effectiveness of the proposed method.

SemTra: A Semantic Skill Translator for Cross-Domain Zero-Shot Policy Adaptation

This work explores the zero-shot adaptation capability of semantic skills, semantically interpretable experts' behavior patterns, in cross-domain settings, where a user input in interleaved multi-modal snippets can prompt a new long-horizon task for different domains. In these cross-domain settings, we present a semantic skill translator framework SemTra which utilizes a set of multi-modal models to extract skills from the snippets, and leverages the reasoning capabilities of a pretrained language model to adapt these extracted skills to the target domain. The framework employs a two-level hierarchy for adaptation: task adaptation and skill adaptation. During task adaptation, seq-to-seq translation by the language model transforms the extracted skills into a semantic skill sequence, which is tailored to fit the cross-domain contexts. Skill adaptation focuses on optimizing each semantic skill for the target domain context, through parametric instantiations that are facilitated by language prompting and contrastive learning-based context inferences. This hierarchical adaptation empowers the framework to not only infer a complex task specification in one-shot from the interleaved multi-modal snippets, but also adapt it to new domains with zero-shot learning abilities. We evaluate our framework with Meta-World, Franka Kitchen, RLBench, and CARLA environments. The results clarify the framework's superiority in performing long-horizon tasks and adapting to different domains, showing its broad applicability in practical use cases, such as cognitive robots interpreting abstract instructions and autonomous vehicles operating under varied configurations.

Revealing the Proximate Long-Tail Distribution in Compositional Zero-Shot Learning

Compositional Zero-Shot Learning (CZSL) aims to transfer knowledge from seen state-object pairs to novel unseen pairs. In this process, visual bias caused by the diverse interrelationship of state-object combinations blurs their visual features, hindering the learning of distinguishable class prototypes. Prevailing methods concentrate on disentangling states and objects directly from visual features, disregarding potential enhancements that could arise from a data viewpoint. Experimentally, we unveil the results caused by the above problem closely approximate the long-tailed distribution. As a solution, we transform CZSL into a proximate class imbalance problem. We mathematically deduce the role of class prior within the long-tailed distribution in CZSL. Building upon this insight, we incorporate visual bias caused by compositions into the classifier's training and inference by estimating it as a proximate class prior. This enhancement encourages the classifier to acquire more discernible class prototypes for each composition, thereby achieving more balanced predictions. Experimental results demonstrate that our approach elevates the model's performance to the state-of-the-art level, without introducing additional parameters.

ProCC: Progressive Cross-Primitive Compatibility for Open-World Compositional Zero-Shot Learning

Open-World Compositional Zero-shot Learning (OW-CZSL) aims to recognize novel compositions of state and object primitives in images with no priors on the compositional space, which induces a tremendously large output space containing all possible state-object compositions. Existing works either learn the joint compositional state-object embedding or predict simple primitives with separate classifiers. However, the former method heavily relies on external word embedding methods, and the latter ignores the interactions of interdependent primitives, respectively. In this paper, we revisit the primitive prediction approach and propose a novel method, termed Progressive Cross-primitive Compatibility (ProCC), to mimic the human learning process for OW-CZSL tasks. Specifically, the cross-primitive compatibility module explicitly learns to model the interactions of state and object features with the trainable memory units, which efficiently acquires cross-primitive visual attention to reason high-feasibility compositions, without the aid of external knowledge. Moreover, to alleviate the invalid cross-primitive interactions, especially for partial-supervision conditions (pCZSL), we design a progressive training paradigm to optimize the primitive classifiers conditioned on pre-trained features in an easy-to-hard manner. Extensive experiments on three widely used benchmark datasets demonstrate that our method outperforms other representative methods on both OW-CZSL and pCZSL settings by large margins.

Automated Category and Trend Analysis of Scientific Articles on Ophthalmology Using Large Language Models: Development and Usability Study.

In this paper, we present an automated method for article classification, leveraging the power of large language models (LLMs). The aim of this study is to evaluate the applicability of various LLMs based on textual content of scientific ophthalmology papers. We developed a model based on natural language processing techniques, including advanced LLMs, to process and analyze the textual content of scientific papers. Specifically, we used zero-shot learning LLMs and compared Bidirectional and Auto-Regressive Transformers (BART) and its variants with Bidirectional Encoder Representations from Transformers (BERT) and its variants, such as distilBERT, SciBERT, PubmedBERT, and BioBERT. To evaluate the LLMs, we compiled a data set (retinal diseases [RenD] ) of 1000 ocular disease-related articles, which were expertly annotated by a panel of 6 specialists into 19 distinct categories. In addition to the classification of articles, we also performed analysis on different classified groups to find the patterns and trends in the field. The classification results demonstrate the effectiveness of LLMs in categorizing a large number of ophthalmology papers without human intervention. The model achieved a mean accuracy of 0.86 and a mean F1-score of 0.85 based on the RenD data set. The proposed framework achieves notable improvements in both accuracy and efficiency. Its application in the domain of ophthalmology showcases its potential for knowledge organization and retrieval. We performed a trend analysis that enables researchers and clinicians to easily categorize and retrieve relevant papers, saving time and effort in literature review and information gathering as well as identification of emerging scientific trends within different disciplines. Moreover, the extendibility of the model to other scientific fields broadens its impact in facilitating research and trend analysis across diverse disciplines.

Ensemble learning with soft-prompted pretrained language models for fact checking

The infectious diseases, such as COVID-19 pandemic, has led to a surge of information on the internet, including misinformation, necessitating fact-checking tools. However, fact-checking infectious diseases related claims pose challenges due to informal claims versus formal evidence and the presence of multiple aspects in a claim. To address these issues, we propose a soft prompt-based ensemble learning framework for COVID-19 fact checking. To understand complex assertions in informal social media texts, we explore various soft prompt structures to take advantage of the T5 language model, and ensemble these prompt structures together. Soft prompts offer flexibility and better generalization compared to hard prompts. The ensemble model captures linguistic cues and contextual information in COVID-19-related data, and thus enhances generalization to new claims. Experimental results demonstrate that prompt-based ensemble learning improves fact-checking accuracy and provides a promising approach to combat misinformation during the pandemic. In addition, the method also shows great zero-shot learning capability and thus can be applied to various fact checking problems.

Zero-Shot Recommendation AI Models for Efficient Job–Candidate Matching in Recruitment Process

In the evolving realities of recruitment, the precision of job–candidate matching is crucial. This study explores the application of Zero-Shot Recommendation AI Models to enhance this matching process. Utilizing advanced pretrained models such as all-MiniLM-L6-v2 and applying similarity metrics like dot product and cosine similarity, we assessed their effectiveness in aligning job descriptions with candidate profiles. Our evaluations, based on Top-K Accuracy across various rankings, revealed a notable enhancement in matching accuracy compared to conventional methods. Specifically, the all-MiniLM-L6-v2 model with a chunk length of 768 exhibited outstanding performance, achieving a remarkable Top-1 accuracy of 3.35%, 55.45% for Top-100, and an impressive 81.11% for Top-500, establishing it as a highly effective tool for recruitment processes. This paper presents an in-depth analysis of these models, providing insights into their potential applications in real-world recruitment scenarios. Our findings highlight the capability of Zero-Shot Learning to address the dynamic requirements of the job market, offering a scalable, efficient, and adaptable solution for job–candidate matching and setting new benchmarks in recruitment efficiency.

Natural language instructions induce compositional generalization in networks of neurons

A fundamental human cognitive feat is to interpret linguistic instructions in order to perform novel tasks without explicit task experience. Yet, the neural computations that might be used to accomplish this remain poorly understood. We use advances in natural language processing to create a neural model of generalization based on linguistic instructions. Models are trained on a set of common psychophysical tasks, and receive instructions embedded by a pretrained language model. Our best models can perform a previously unseen task with an average performance of 83% correct based solely on linguistic instructions (that is, zero-shot learning). We found that language scaffolds sensorimotor representations such that activity for interrelated tasks shares a common geometry with the semantic representations of instructions, allowing language to cue the proper composition of practiced skills in unseen settings. We show how this model generates a linguistic description of a novel task it has identified using only motor feedback, which can subsequently guide a partner model to perform the task. Our models offer several experimentally testable predictions outlining how linguistic information must be represented to facilitate flexible and general cognition in the human brain.

Dual-channel representation consistent recommender for session-based new item recommendation

Session-Based Recommendations (SBR) have become a crucial branch in the recommendation field. Usually, most previous SBR methods only recommend items existing in users’ historical sessions, facing difficulties with uninteracted items. Introducing new items to users is essential, but solving this issue termed as “Session-based New Item Recommendation (SNIR)” is not an easy task. The complexity arises from accurately characterizing new items and predicting user preferences without user-item interactions. To tackle these challenges, we propose a simple yet effective model named Dual-Channel Representation Consistent Recommender (DC2R). We design a dual-channel strategy to effectively model a new item by its ontology information and inferred representation. Specifically, the ontology information is captured by integrating the inherent attributes of the item into the learning. Meanwhile, leveraging the zero-shot learning technique, we can obtain the inferred item representation from additional attributes. Dual preferences are extracted from the two representations separately to avoid preference bias issues. To better mine user preferences, a Representation Consistent Encoder is utilized to ensure that user preferences and item representations are in the same linear space, thus more precisely reflecting users’ interests. Experiments are conducted on three representative real-world datasets, and the results demonstrate the effectiveness of our proposed method. The source code of our proposed model is available at https://github.com/codeCyw/DC2R.

Improving Automated Program Repair with Domain Adaptation

Automated Program Repair (APR) is defined as the process of fixing a bug/defect in the source code, by an automated tool. APR tools have recently experienced promising results by leveraging state-of-the-art Neural Language Processing (NLP) techniques. APR tools such as TFix and CodeXGLUE that combine text-to-text transformers with software-specific techniques are outperforming alternatives, these days. However, in most APR studies, the train and test sets are chosen from the same set of projects (i.e., when APR fixes a bug in the test set from project A, the model has already seen example fixed bugs from project A in the training set). In the real world, however, APR models are meant to be generalizable to new and different projects. Therefore, there is a potential threat that reported APR models with high effectiveness perform poorly when the characteristics of the new project or its bugs are different than the training set’s (“Domain Shift”). In this study, we first define the problem of domain shift in automated program repair. Next, we measure the potential damage of domain shift on two recent APR models (TFix and CodeXGLUE). Based on this observation, we then propose a domain adaptation framework that can adapt an APR model for a given target project. We conduct an empirical study with three domain adaptation methods FullFineTuning , TuningWithLightWeightAdapterLayers , and CurriculumLearning and two APR models on 2,672 bugs from 12 projects. The results show that our proposed framework on average can improve the effectiveness of TFix by 13.05% and CodeXGLUE by 48.78%, in terms of “Exact Match”. Through experiments, we also show that the framework provides high efficiency and reliability (in terms of “Exposure Bias”). Using synthetic data to domain adapt TFix and CodeXGLUE on the projects with no data (Zero-shot learning), also results in an average improvement of 5.76% and 17.62% for TFix and CodeXGLUE, respectively.

Transfer and zero-shot learning for scalable weed detection and classification in UAV images

In an effort to reduce pesticide use, agronomists and computer scientists have joined forces to develop site-specific weed detection and classification systems. These systems aim to recognize and locate weed species within a crop field, using precision equipment to apply required herbicides timely and only where needed, with the objective of reducing the sprayable surface required to eliminate the given weed and protect the crop, with both economic and environmental benefits. Yet, with climate change on the rise, common weeds are expected to undergo some changes to adapt to their environment, possibly with new or invasive weeds spreading to areas where they did not exist before. These changes (often morphological) as well as new invasions need to be taken into account by future classifiers and detection algorithms to ensure system robustness and adaptation to new habitats/climate dynamics. This paper proposes a set of experiments evaluating the use of transfer learning and zero-shot learning for weed classification using our novel TomatoWeeds dataset. Residual networks of variable depth, pretrained on the Imagenet and/or DeepWeeds datasets were evaluated. A ResNet50 pretrained on both datasets and fine-tuned on the TomatoWeeds dataset performed best, returning a holdout set accuracy of 77.8%, showing the advantageous use of transfer learning in this domain. Zero-shot learning, using both embeddings of images and morphological and habitat text-based descriptions, is implemented to test the ability of machine learning pipelines of recognizing unseen classes at test time (which may arise e.g. due to changing climate dynamics), a learning task in which the field (and our experiments) are still far from satisfactory results. Further research could benefit from larger weed-specific datasets for transfer learning as well as deeper network architectures to improve model performance. The projection-based ZSL could also benefit from larger datasets and new zero-shot learning architectures in hope that unseen classes are accurately projected.

Artificial Intelligence Methods in Natural Language Processing: A Comprehensive Review

The rapid evolution of Artificial Intelligence (AI) since its inception in the mid-20th century has significantly influenced the field of Natural Language Processing (NLP), transforming it from a rule-based system to a dynamic and adaptive model capable of understanding the complexities of human language. This paper aims to offer a comprehensive review of the various applications and methodologies of AI in NLP, serving as a detailed guide for future research and practical applications. In the early sections, the paper elucidates the indispensable role of AI in NLP, highlighting its transition from symbolic reasoning to a focus on machine learning and deep learning, and its extensive applications in sectors such as healthcare, transportation, and finance. It emphasizes the symbiotic relationship between AI and NLP, facilitated by platforms like AllenNLP, which aid in the development of advanced language understanding models. Further, the paper explores specific AI techniques employed in NLP, including machine learning, Naive Bayes, and Support Vector Machines, and identifies pressing challenges and avenues for future research. It delves into the applications of AI in NLP, showcasing its transformative potential in tasks such as machine translation, facilitated by deep learning methods, and the development of chatbots and virtual assistants that have revolutionized human-technology interaction. The paper also highlights other fields impacted by AI techniques, including text summarization, sentiment analysis, and named entity recognition, emphasizing the efficiency and accuracy brought about by the integration of AI in these areas. In conclusion, the paper summarizes the remarkable advancements and persistent challenges in NLP, such as language ambiguity and contextual understanding, and underscores the need for diverse and representative labeled data for training. Looking forward, it identifies promising research avenues including Explainable AI, Few-shot and Zero-shot Learning, and the integration of NLP with other data modalities, aiming for a holistic understanding of multimodal data. The paper calls for enhanced robustness and security in NLP systems, especially in sensitive applications like content moderation and fake news detection, to foster trust and reliability in AI technologies. It advocates for continual learning in NLP models to adapt over time without losing previously acquired knowledge, paving the way for a future where AI and NLP work synergistically to understand and generate human language more effectively and efficiently.

A platform-independent AI tumor lineage and site (ATLAS) classifier

Histopathologic diagnosis and classification of cancer plays a critical role in guiding treatment. Advances in next-generation sequencing have ushered in new complementary molecular frameworks. However, existing approaches do not independently assess both site-of-origin (e.g. prostate) and lineage (e.g. adenocarcinoma) and have minimal validation in metastatic disease, where classification is more difficult. Utilizing gradient-boosted machine learning, we developed ATLAS, a pair of separate AI Tumor Lineage and Site-of-origin models from RNA expression data on 8249 tumor samples. We assessed performance independently in 10,376 total tumor samples, including 1490 metastatic samples, achieving an accuracy of 91.4% for cancer site-of-origin and 97.1% for cancer lineage. High confidence predictions (encompassing the majority of cases) were accurate 98–99% of the time in both localized and remarkably even in metastatic samples. We also identified emergent properties of our lineage scores for tumor types on which the model was never trained (zero-shot learning). Adenocarcinoma/sarcoma lineage scores differentiated epithelioid from biphasic/sarcomatoid mesothelioma. Also, predicted lineage de-differentiation identified neuroendocrine/small cell tumors and was associated with poor outcomes across tumor types. Our platform-independent single-sample approach can be easily translated to existing RNA-seq platforms. ATLAS can complement and guide traditional histopathologic assessment in challenging situations and tumors of unknown primary.

Co-GZSL: Feature Contrastive Optimization for Generalized Zero-Shot Learning

Generalized Zero-Shot Learning (GZSL) learns from only labeled seen classes during training but discriminates both seen and unseen classes during testing. In GZSL tasks, most of the existing methods commonly utilize visual and semantic features for training. Due to the lack of visual features for unseen classes, recent works generate real-like visual features by using semantic features. However, the synthesized features in the original feature space lack discriminative information. It is important that the synthesized visual features should be similar to the ones in the same class, but different from the other classes. One way to solve this problem is to introduce the embedding space after generating visual features. Following this situation, the embedded features from the embedding space can be inconsistent with the original semantic features. For another way, some recent methods constrain the representation by reconstructing the semantic features using the original visual features and the synthesized visual features. In this paper, we propose a hybrid GZSL model, named feature Contrastive optimization for GZSL (Co-GZSL), to reconstruct the semantic features from the embedded features, which ensures that the embedded features are close to the original semantic features indirectly by comparing reconstructed semantic features with original semantic features. In addition, to settle the problem that the synthesized features lack discrimination and semantic consistency, we introduce a Feature Contrastive Optimization Module (FCOM) and jointly utilize contrastive and semantic cycle-consistency losses in the FCOM to strengthen the intra-class compactness and the inter-class separability and to encourage the model to generate semantically consistent and discriminative visual features. By combining the generative module, the embedding module, and the FCOM, we achieve Co-GZSL. We evaluate the proposed Co-GZSL model on four benchmarks, and the experimental results indicate that our model is superior over current methods. Code is available at: https://github.com/zhanzhuxi/Co-GZSL.

Target Detection and Segmentation Technology for Zero-shot Learning

Zero-shot learning (ZSL) in the field of computer vision refers to enabling the model to recognize and understand categories that have not been encountered during the training phase. It is particularly critical for object detection and segmentation tasks, because these tasks require the model to have good generalization capabilities to unknown categories. Object detection requires the model to determine the location of the object, while segmentation further requires the precise demarcation of the object's boundaries. In ZSL research, knowledge representation and transfer are core issues. Researchers have tried to use semantic attributes as a knowledge bridge to connect categories seen during the training phase and categories not seen during the testing phase. These attributes may be color, shape, etc., but this method requires accurate attribute annotation, which is often not easy to achieve in practice. Therefore, researchers have begun to explore the use of non-visual information such as knowledge maps and text descriptions to enrich the recognition capabilities of models, but this also introduces the challenge of information integration and alignment. At present, ZSL has made certain progress in target detection and segmentation tasks, but there is still a significant gap compared with traditional supervised learning. This is mainly due to the limited ability of ZSL models to generalize to new categories. To this end, researchers have begun to explore combining ZSL with other technologies, such as generative adversarial networks (GANs) and reinforcement learning, to enhance the model's detection and segmentation capabilities for new categories. Future research needs to focus on several aspects. The first is how to design a more effective knowledge representation and transfer mechanism so that the model can better utilize existing knowledge. The second step is to develop new algorithms to improve the performance of ZSL in complex environments. In addition, research should focus on how to reduce the dependence on computing resources so that the ZSL method can run effectively in resource-limited environments. In summary, the research on target detection and segmentation technology of zero-shot learning is a cutting-edge topic in the field of computer vision. Despite the challenges, with the deepening of research, we expect these technologies to contribute to improving the generalization ability and intelligence level of computer vision systems.

Adaptive Relation-Aware Network for zero-shot classification

Supervised learning-based image classification in computer vision relies on visual samples containing a large amount of labeled information. Considering that it is labor-intensive to collect and label images and construct datasets manually, Zero-Shot Learning (ZSL) achieves knowledge transfer from seen categories to unseen categories by mining auxiliary information, which reduces the dependence on labeled image samples and is one of the current research hotspots in computer vision. However, most ZSL methods fail to properly measure the relationships between classes, or do not consider the differences and similarities between classes at all. In this paper, we propose Adaptive Relation-Aware Network (ARAN), a novel ZSL approach that incorporates the improved triplet loss from deep metric learning into a VAE-based generative model, which helps to model inter-class and intra-class relationships for different classes in ZSL datasets and generate an arbitrary amount of high-quality visual features containing more discriminative information. Moreover, we validate the effectiveness and superior performance of our ARAN through experimental evaluations under ZSL and more practical GZSL settings on three popular datasets AWA2, CUB, and SUN.

Data extraction for evidence synthesis using a large language model: A proof-of-concept study.

Data extraction is a crucial, yet labor-intensive and error-prone part of evidence synthesis. To date, efforts to harness machine learning for enhancing efficiency of the data extraction process have fallen short of achieving sufficient accuracy and usability. With the release of large language models (LLMs), new possibilities have emerged to increase efficiency and accuracy of data extraction for evidence synthesis. The objective of this proof-of-concept study was to assess the performance of an LLM (Claude 2) in extracting data elements from published studies, compared with human data extraction as employed in systematic reviews. Our analysis utilized a convenience sample of 10 English-language, open-access publications of randomized controlled trials included in a single systematic review. We selected 16 distinct types of data, posing varying degrees of difficulty (160 data elements across 10 studies). We used the browser version of Claude 2 to upload the portable document format of each publication and then prompted the model for each data element. Across 160 data elements, Claude 2 demonstrated an overall accuracy of 96.3% with a high test-retest reliability (replication 1: 96.9%; replication 2: 95.0% accuracy). Overall, Claude 2 made 6 errors on 160 data items. The most common errors (n = 4) were missed data items. Importantly, Claude 2's ease of use was high; it required no technical expertise or labeled training data for effective operation (i.e., zero-shot learning). Based on findings of our proof-of-concept study, leveraging LLMs has the potential to substantially enhance the efficiency and accuracy of data extraction for evidence syntheses.

Differential Refinement Network for Zero-Shot Learning.

Zero-shot learning (ZSL) aims to recognize novel categories by merely utilizing disjoint seen samples. It is a challenging task as the knowledge of unseen objects is forbidden in the training stage, which easily leads to unseen samples degrading to mismatched categories. In order to alleviate the biased recognition problem, in this article, we propose a differential refinement network (DRNet) for ZSL, which aims to explore robust semantic-to-visual embedding. Our DRNet model consists of two subnetworks: basic network and differential network. The basic network targets to generate initial class-specific visual centers conditioned on corresponding semantic prototypes. The differential network is designed to predict class-unrelated differences between visual centers of arbitrary semantic prototype pairs, which are applied to further polish the initial visual centers. The motivation is that, by comparing different prototypes, interactions between various categories will be characterized, benefiting the generation of authentic and discriminative visual centers. Moreover, a modified episode-based training paradigm is explored to optimize the two subnetworks actively. In the training stage, we form a collection of episodes, each of which is an imitated ZSL task. Our DRNet is optimized by those sampled tasks rather than individual samples, which progressively learns skills to adapt and generalize to novel classes. Experiments on four challenging datasets demonstrate the effectiveness of our method.

Zero-Shot Learning With Attentive Region Embedding and Enhanced Semantics.

The performance of zero-shot learning (ZSL) can be improved progressively by learning better features and generating pseudosamples for unseen classes. Existing ZSL works typically learn feature extractors and generators independently, which may shift the unseen samples away from their real distribution and suffers from the domain bias problem. In this article, to tackle this challenge, we propose a variational autoencoder (VAE)-based framework, that is, joint Attentive Region Embedding with Enhanced Semantics (AREES), which is tailored to advance the zero-shot recognition. Specifically, AREES is end-to-end trainable and consists of three network branches: 1) attentive region embedding is used to learn the semantic-guided visual features by the attention mechanism (AM); 2) a decomposition structure and a semantic pivot regularization are used to extract enhanced semantics; and 3) a multimodal VAE (mVAE) with the cross-reconstruction loss and the distribution alignment loss is used to obtain a shared latent embedding space of visual features and semantics. Finally, features' extraction and features' generation are optimized together in AREES to address the domain shift problem to a large extent. The comprehensive evaluations on six benchmarks, including the ImageNet, demonstrate the superiority of the proposed model over its state-of-the-art counterparts.

Zero-Shot Medical Image Translation via Frequency-Guided Diffusion Models.

Recently, the diffusion model has emerged as a superior generative model that can produce high quality and realistic images. However, for medical image translation, the existing diffusion models are deficient in accurately retaining structural information since the structure details of source domain images are lost during the forward diffusion process and cannot be fully recovered through learned reverse diffusion, while the integrity of anatomical structures is extremely important in medical images. For instance, errors in image translation may distort, shift, or even remove structures and tumors, leading to incorrect diagnosis and inadequate treatments. Training and conditioning diffusion models using paired source and target images with matching anatomy can help. However, such paired data are very difficult and costly to obtain, and may also reduce the robustness of the developed model to out-of-distribution testing data. We propose a frequency-guided diffusion model (FGDM) that employs frequency-domain filters to guide the diffusion model for structure-preserving image translation. Based on its design, FGDM allows zero-shot learning, as it can be trained solely on the data from the target domain, and used directly for source-to-target domain translation without any exposure to the source-domain data during training. We evaluated it on three cone-beam CT (CBCT)-to-CT translation tasks for different anatomical sites, and a cross-institutional MR imaging translation task. FGDM outperformed the state-of-the-art methods (GAN-based, VAE-based, and diffusion-based) in metrics of Fréchet Inception Distance (FID), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM), showing its significant advantages in zero-shot medical image translation.