Author name cluster

Dong Hoon Lee

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

8 papers

1 author row

NeurIPS Conference 2025 Conference Paper

Disentangled Representation Learning via Modular Compositional Bias

whie jung
Dong Hoon Lee
Seunghoon Hong

Recent disentangled representation learning (DRL) methods heavily rely on factor-specific strategies—either learning objectives for attributes or model architectures for objects—to embed inductive biases. Such divergent approaches result in significant overhead when novel factors of variation do not align with prior assumptions, such as statistical independence or spatial exclusivity, or when multiple factors coexist, as practitioners must redesign architectures or objectives. To address this, we propose a compositional bias, a modular inductive bias decoupled from both objectives and architectures. Our key insight is that different factors obey distinct "recombination rules" in the data distribution: global attributes are mutually exclusive, e. g. , a face has one nose, while objects share a common support (any subset of objects can co-exist). We therefore randomly remix latents according to factor-specific rules, i. e. , a mixing strategy, and force the encoder to discover whichever factor structure the mixing strategy reflects through two complementary objectives: (i) a prior loss that ensures every remix decodes into a realistic image, and (ii) the compositional consistency loss introduced by Wiedemer et al. , which aligns each composite image with its corresponding composite latent. Under this general framework, simply adjusting the mixing strategy enables disentanglement of attributes, objects, and even both, without modifying the objectives or architectures. Extensive experiments demonstrate that our method shows competitive performance in both attribute and object disentanglement, and uniquely achieves joint disentanglement of global style and objects. Code is available at https: //github. com/whieya/Compositional-DRL.

PDF Details

NeurIPS Conference 2025 Conference Paper

Universal Few-shot Spatial Control for Diffusion Models

Kiet Nguyen
Chanhyuk Lee
Donggyun Kim
Dong Hoon Lee
Seunghoon Hong

Spatial conditioning in pretrained text-to-image diffusion models has significantly improved fine-grained control over the structure of generated images. However, existing control adapters exhibit limited adaptability and incur high training costs when encountering novel spatial control conditions that differ substantially from the training tasks. To address this limitation, we propose Universal Few-Shot Control (UFC), a versatile few-shot control adapter capable of generalizing to novel spatial conditions. Given a few image-condition pairs of an unseen task and a query condition, UFC leverages the analogy between query and support conditions to construct task-specific control features, instantiated by a matching mechanism and an update on a small set of task-specific parameters. Experiments on six novel spatial control tasks show that UFC, fine-tuned with only 30 annotated examples of novel tasks, achieves fine-grained control consistent with the spatial conditions. Notably, when fine-tuned with 0. 1\% of the full training data, UFC achieves competitive performance with the fully supervised baselines in various control tasks. We also show that UFC is applicable agnostically to various diffusion backbones and demonstrate its effectiveness on both UNet and DiT architectures. Code is available at https: //github. com/kietngt00/UFC.

PDF Details

NeurIPS Conference 2024 Conference Paper

Learning to Merge Tokens via Decoupled Embedding for Efficient Vision Transformers

Dong Hoon Lee
Seunghoon Hong

Recent token reduction methods for Vision Transformers (ViTs) incorporate token merging, which measures the similarities between token embeddings and combines the most similar pairs. However, their merging policies are directly dependent on intermediate features in ViTs, which prevents exploiting features tailored for merging and requires end-to-end training to improve token merging. In this paper, we propose Decoupled Token Embedding for Merging (DTEM) that enhances token merging through a decoupled embedding learned via a continuously relaxed token merging process. Our method introduces a lightweight embedding module decoupled from the ViT forward pass to extract dedicated features for token merging, thereby addressing the restriction from using intermediate features. The continuously relaxed token merging, applied during training, enables us to learn the decoupled embeddings in a differentiable manner. Thanks to the decoupled structure, our method can be seamlessly integrated into existing ViT backbones and trained either modularly by learning only the decoupled embeddings or end-to-end by fine-tuning. We demonstrate the applicability of DTEM on various tasks, including classification, captioning, and segmentation, with consistent improvement in token merging. Especially in the ImageNet-1k classification, DTEM achieves a 37. 2\% reduction in FLOPs while maintaining a top-1 accuracy of 79. 85\% with DeiT-small.

PDF Details DOI

NeurIPS Conference 2022 Conference Paper

Unsupervised Visual Representation Learning via Mutual Information Regularized Assignment

Dong Hoon Lee
Sungik Choi
Hyunwoo J. Kim
Sae-Young Chung

This paper proposes Mutual Information Regularized Assignment (MIRA), a pseudo-labeling algorithm for unsupervised representation learning inspired by information maximization. We formulate online pseudo-labeling as an optimization problem to find pseudo-labels that maximize the mutual information between the label and data while being close to a given model probability. We derive a fixed-point iteration method and prove its convergence to the optimal solution. In contrast to baselines, MIRA combined with pseudo-label prediction enables a simple yet effective clustering-based representation learning without incorporating extra training techniques or artificial constraints such as sampling strategy, equipartition constraints, etc. With relatively small training epochs, representation learned by MIRA achieves state-of-the-art performance on various downstream tasks, including the linear/${\it k}$-NN evaluation and transfer learning. Especially, with only 400 epochs, our method applied to ImageNet dataset with ResNet-50 architecture achieves 75. 6% linear evaluation accuracy.

PDF Details

TCS Journal 2020 Journal Article

Tight security for the generic construction of identity-based signature (in the multi-instance setting)

Youngkyung Lee
Jong Hwan Park
Kwangsu Lee
Dong Hoon Lee

An identity-based signature (IBS) scheme can be generically constructed from any ordinary signature scheme by appending a chain of signatures. Until now, it has been known that a generic construction cannot lead to a tightly secure IBS scheme, although any tightly secure signature scheme can be used as a building block. In this study, we demonstrate that the generic construction of IBS can achieve tightness if the underlying signature scheme is tightly secure in the multi-user setting with corruption. In addition, we extend the tightness result of IBS to the multi-instance setting, where an adversary can corrupt multiple key generation centers and obtain multiple related master secret keys. As instantiations, we adopt the efficient and tightly secure signature scheme in the multi-user setting with corruption, recently proposed by Gjøsteen and Jager (CRYPTO 2018). As a result, we can obtain the first efficient and tightly secure IBS schemes (in the multi-instance setting) based on the Diffie–Hellman assumptions in the random oracle model.

Details DOI

TCS Journal 2017 Journal Article

Self-updatable encryption: Time constrained access control with hidden attributes and better efficiency

Kwangsu Lee
Seung Geol Choi
Dong Hoon Lee
Jong Hwan Park
Moti Yung

Revocation and key evolving paradigms are central issues in cryptography, and in PKI in particular. A novel concern related to these areas was raised in the recent work of Sahai, Seyalioglu, and Waters (CRYPTO 2012) who noticed that revoking past keys should at times (e. g. , the scenario of cloud storage) be accompanied by revocation of past ciphertexts (to prevent unread ciphertexts from being read by revoked users). They introduced revocable-storage attribute-based encryption (RS-ABE) as a good access control mechanism for cloud storage. RS-ABE protects against the revoked users not only the future data by supporting key-revocation but also the past data by supporting ciphertext-update, through which a ciphertext at time T can be updated to a new ciphertext at time T + 1 using only the public key. Motivated by this pioneering work, we ask whether it is possible to have a modular approach, which includes a primitive for time managed ciphertext update as a primitive. We call encryption which supports this primitive a “self-updatable encryption” (SUE). We then suggest a modular cryptosystems design methodology based on three sub-components: a primary encryption scheme, a key-revocation mechanism, and a time-evolution mechanism which controls the ciphertext self-updating via an SUE method, coordinated with the revocation (when needed). Our goal in this is to allow the self-updating ciphertext component to take part in the design of new and improved cryptosystems and protocols in a flexible fashion. Specifically, we achieve the following results: We first introduce a new cryptographic primitive called self-updatable encryption (SUE), realizing a time-evolution mechanism. In SUE, a ciphertext and a private key are associated with time. A user can decrypt a ciphertext if its time is earlier than that of his private key. Additionally, anyone (e. g. , a cloud server) can update the ciphertext to a ciphertext with a newer time. We also construct an SUE scheme and prove its full security under static assumptions. Following our modular approach, we present a new RS-ABE scheme with shorter ciphertexts than that of Sahai et al. and prove its security. The length efficiency is mainly due to our SUE scheme and the underlying modularity. We apply our approach to predicate encryption (PE) supporting attribute-hiding property, and obtain a revocable-storage PE (RS-PE) scheme that is selectively-secure. We further demonstrate that SUE is of independent interest, by showing it can be used for timed-release encryption (and its applications), and for augmenting key-insulated encryption with forward-secure storage.

Details DOI

TCS Journal 2015 Journal Article

Sequential aggregate signatures with short public keys without random oracles

Kwangsu Lee
Dong Hoon Lee
Moti Yung

The notion of aggregate signature has been motivated by applications and it enables any user to compress different signatures signed by different signers on different messages into a short signature. Sequential aggregate signature, in turn, is a special kind of aggregate signature that only allows a signer to add his signature into an aggregate signature in sequential order. This latter scheme has applications in diversified settings such as in reducing bandwidth of certificate chains and in secure routing protocols. Lu, Ostrovsky, Sahai, Shacham, and Waters (EUROCRYPT 2006) presented the first sequential aggregate signature scheme in the standard model. The size of their public key, however, is quite large (i. e. , the number of group elements is proportional to the security parameter), and therefore, they suggested as an open problem the construction of such a scheme with short keys. In this paper, we propose the first sequential aggregate signature schemes with short public keys (i. e. , a constant number of group elements) in prime order (asymmetric) bilinear groups that are secure under static assumptions in the standard model. Furthermore, our schemes employ a constant number of pairing operations per message signing and message verification operation. Technically, we start with a public-key signature scheme based on the recent dual system encryption technique of Lewko and Waters (TCC 2010). This technique cannot directly provide an aggregate signature scheme since, as we observed, additional elements should be published in a public key to support aggregation. Thus, our constructions are careful augmentation techniques for the dual system technique to allow it to support sequential aggregate signature schemes. We also propose a multi-signature scheme with short public parameters in the standard model.

Details DOI

TCS Journal 1993 Journal Article

Separating k-separated eNCE graph languages

Changwook Kim
Dong Hoon Lee

An eNCE graph grammar is k-separated (k⩾1) if the distance between any two nonterminal nodes in any of its sentential forms is at least k. Let SEP k denote the class of graph languages generated by k-separated grammars. Then, SEP1 (SEP2) is the class of eNCE (boundary eNCE) graph languages, and so SEP2⊊SEP1. Recently, Engelfriet (1991) showed that SEP3⊊SEP2 and conjectured that, in fact, SEP k+1 ⊊SEP k for each k⩾ 1. We prove this conjecture affirmatively.

Details DOI