arXiv subjects

Jinho Chang

Publications and source records attributed to Jinho Chang.

14 recordsLinked to original sources

arXiv2026

Reward Score Matching: Unifying Reward-based Fine-tuning for Flow and Diffusion Models

Reward-based fine-tuning steers a pretrained diffusion or flow-based generative model toward higher-reward samples while remaining close to the pretrained model. Although existing methods are derived from different perspectives, we show that many can be written under a common framework, which we call reward score matching (RSM). Under this view, alignment becomes score matching against a value-guided target, and the main differences across methods reduce to the construction of the value-guidance estimator and the effective optimization strength across timesteps. This unification clarifies the bias-variance-compute tradeoffs of existing designs, and distinguishes core optimization components from auxiliary mechanisms that add complexity without clear benefit. Guided by this perspective, we develop simpler, more efficient redesigns across representative differentiable and black-box reward alignment tasks. Overall, RSM turns a seemingly fragmented collection of reward-based fine-tuning methods into a smaller, more interpretable, and more actionable design space. Code is available at https://github.com/jaylee2000/rsm.

cs.LG↗

arXiv2025

Training-Free Reward-Guided Image Editing via Trajectory Optimal Control

Recent advancements in diffusion and flow-matching models have demonstrated remarkable capabilities in high-fidelity image synthesis. A prominent line of research involves reward-guided guidance, which steers the generation process during inference to align with specific objectives. However, leveraging this reward-guided approach to the task of image editing, which requires preserving the semantic content of the source image while enhancing a target reward, is largely unexplored. In this work, we introduce a novel framework for training-free, reward-guided image editing. We formulate the editing process as a trajectory optimal control problem where the reverse process of a diffusion model is treated as a controllable trajectory originating from the source image, and the adjoint states are iteratively updated to steer the editing process. Through extensive experiments across distinct editing tasks, we demonstrate that our approach significantly outperforms existing inversion-based training-free guidance baselines, achieving a superior balance between reward maximization and fidelity to the source image without reward hacking.

cs.CV↗

arXiv2024

ContrastiveCFG: Guiding Diffusion Sampling by Contrasting Positive and Negative Concepts

As Classifier-Free Guidance (CFG) has proven effective in conditional diffusion model sampling for improved condition alignment, many applications use a negated CFG term as a Negative Prompting (NP) to filter out unwanted features from samples. However, simply negating CFG guidance creates an inverted probability distribution, often distorting samples away from the marginal distribution. Inspired by recent advances in conditional diffusion models for inverse problems, here we present a novel method to achieve guidance toward the given condition using contrastive loss. Specifically, our guidance term aligns or repels the denoising direction based on the given condition through contrastive loss, achieving a similar guiding effect to traditional CFG for positive conditions while overcoming the limitations of existing negative guidance methods. Experimental results demonstrate that our approach effectively injects or removes the given concepts while maintaining sample quality across diverse scenarios, from simple class conditions to complex and overlapping text prompts.

cs.LG↗

arXiv2024

LDMol: A Text-to-Molecule Diffusion Model with Structurally Informative Latent Space Surpasses AR Models

With the emergence of diffusion models as a frontline generative model, many researchers have proposed molecule generation techniques with conditional diffusion models. However, the unavoidable discreteness of a molecule makes it difficult for a diffusion model to connect raw data with highly complex conditions like natural language. To address this, here we present a novel latent diffusion model dubbed LDMol for text-conditioned molecule generation. By recognizing that the suitable latent space design is the key to the diffusion model performance, we employ a contrastive learning strategy to extract novel feature space from text data that embeds the unique characteristics of the molecule structure. Experiments show that LDMol outperforms the existing autoregressive baselines on the text-to-molecule generation benchmark, being one of the first diffusion models that outperforms autoregressive models in textual data generation with a better choice of the latent domain. Furthermore, we show that LDMol can be applied to downstream tasks such as molecule-to-text retrieval and text-guided molecule editing, demonstrating its versatility as a diffusion model.

cs.LG↗

arXiv2024

Ground-A-Score: Scaling Up the Score Distillation for Multi-Attribute Editing

Despite recent advancements in text-to-image diffusion models facilitating various image editing techniques, complex text prompts often lead to an oversight of some requests due to a bottleneck in processing text information. To tackle this challenge, we present Ground-A-Score, a simple yet powerful model-agnostic image editing method by incorporating grounding during score distillation. This approach ensures a precise reflection of intricate prompt requirements in the editing outcomes, taking into account the prior knowledge of the object locations within the image. Moreover, the selective application with a new penalty coefficient and contrastive loss helps to precisely target editing areas while preserving the integrity of the objects in the source image. Both qualitative assessments and quantitative analyses confirm that Ground-A-Score successfully adheres to the intricate details of extended and multifaceted prompts, ensuring high-quality outcomes that respect the original image attributes.

cs.CV↗

arXiv2024

DreamMotion: Space-Time Self-Similar Score Distillation for Zero-Shot Video Editing

Text-driven diffusion-based video editing presents a unique challenge not encountered in image editing literature: establishing real-world motion. Unlike existing video editing approaches, here we focus on score distillation sampling to circumvent the standard reverse diffusion process and initiate optimization from videos that already exhibit natural motion. Our analysis reveals that while video score distillation can effectively introduce new content indicated by target text, it can also cause significant structure and motion deviation. To counteract this, we propose to match space-time self-similarities of the original video and the edited video during the score distillation. Thanks to the use of score distillation, our approach is model-agnostic, which can be applied for both cascaded and non-cascaded video diffusion frameworks. Through extensive comparisons with leading methods, our approach demonstrates its superiority in altering appearances while accurately preserving the original structure and motion.

cs.CV↗

arXiv2023

LLM-CXR: Instruction-Finetuned LLM for CXR Image Understanding and Generation

Following the impressive development of LLMs, vision-language alignment in LLMs is actively being researched to enable multimodal reasoning and visual IO. This direction of research is particularly relevant to medical imaging because medical image analysis and generation consist of reasoning based on a combination of visual features and prior knowledge. Many recent works have focused on training adapter networks that serve as an information bridge between image processing networks and LLMs; but presumably, in order to achieve maximum reasoning potential of LLMs on visual information as well, visual and language features should be allowed to interact more freely. This is especially important in the medical domain because understanding and generating medical images such as chest X-rays (CXR) require not only accurate visual and language-based reasoning but also a more intimate mapping between the two modalities. Thus, taking inspiration from previous work on the transformer and VQ-GAN combination for bidirectional image and text generation, we build upon this approach and develop a method for instruction-tuning an LLM pre-trained only on text to gain vision-language capabilities for medical images. Specifically, we leverage a pretrained LLM's existing question-answering and instruction-following abilities to teach it to understand visual inputs by instructing it to answer questions about image inputs and, symmetrically, output both text and image responses appropriate to a given query by tuning the LLM with diverse tasks that encompass image-based text-generation and text-based image-generation. We show that our model, LLM-CXR, trained in this approach shows better image-text alignment in both CXR understanding and generation tasks while being smaller in size compared to previously developed models that perform a narrower range of tasks. The code is at https://github.com/hyn2028/llm-cxr.

cs.CV↗

arXiv2022

Bidirectional Generation of Structure and Properties Through a Single Molecular Foundation Model

The recent success of large foundation models in artificial intelligence has prompted the emergence of chemical pre-trained models. Despite the growing interest in large molecular pre-trained models that provide informative representations for downstream tasks, attempts for multimodal pre-training approaches on the molecule domain were limited. To address this, we present a novel multimodal molecular pre-trained model that incorporates the modalities of structure and biochemical properties, drawing inspiration from recent advances in multimodal learning techniques. Our proposed model pipeline of data handling and training objectives aligns the structure/property features in a common embedding space, which enables the model to regard bidirectional information between the molecules' structure and properties. These contributions emerge synergistic knowledge, allowing us to tackle both multimodal and unimodal downstream tasks through a single model. Through extensive experiments, we demonstrate that our model shows remarkable capabilities in solving various meaningful chemical challenges, including conditional molecule generation, property prediction, molecule classification, and reaction prediction.

cs.LG↗

arXiv2017

Distilling Entanglement with Noisy Operations

Entanglement distillation is a fundamental task in quantum information processing. It not only extracts entanglement out of corrupted systems but also leads to protecting systems of interest against intervention with environment. In this work, we consider a realistic scenario of entanglement distillation where noisy quantum operations are applied. In particular, the two-way distillation protocol that tolerates the highest error rate is considered. We show that among all types of noise there are only four equivalence classes according to the distillability condition. Since the four classes are connected by local unitary transformations, our results can be used to improve entanglement distillability in practice when entanglement distillation is performed in a realistic setting.

quant-ph↗

arXiv2014

Revisiting Quantum discord for two-qubit X states: Error bound to Analytical formula

In this article, we investigate the error bound of quantum discord, obtained by the analytic formula of Ali et al.[Phys. Rev. A 81(2010), 042105] in case of general X states and by the analytic formula of Fanchini et al.[Phys. Rev. A 81(2010), 052107] in case of symmetric X states. We show that results of Ali et. al. to general X states and Fanchini et al. to symmetric X states may have worst-case error of 0.004565 and 0.0009 respectively.

quant-ph↗

arXiv2012

Geometric discord of quantum states of fermionic system in accelerated frame

In this article, we investigate the geometric discord of quantum states of fermionic system in accelerated frame. It is shown by the method beyond the single-mode approximation, depending on the region considered, that the geometric discord for the entangled quantum states of fermionic system in accelerated frame can vanish or be retained at the infinite acceleration limit: it does not disappear when the quantum state of the particle(Alice)-particle(Bob in region I) case or the particle(Alice)-antiparticle(Bob in region II) is considered and it disappears when the particle(Alice)-antiparticle(Bob in region I) case or the particle(Alice)-particle(Bob in region II) one is considered.

quant-ph↗

arXiv2012

Communication channel of fermionic system in accelerated frame

In this article, we investigate the communication channel of fermionic system in an accelerated frame. We observe that at the infinite acceleration, the mutual information of single rail quantum channel coincides with that of double rail quantum channel, but those of classical ones reach different values. Furthermore, we find that at the infinite acceleration, the conditional entropy of single(or double) rail quantum channel vanishes, but those of classical ones may have finite values. In addition, we see that even when considering a method beyond the single mode approximation, the dual rail entangled state seems to provide better channel capacity than the single rail entangled state, unlike the bosonic case. Moreover, we find that the single-mode approximation is not sufficient to analyze the communication channel of fermionic system in an accelerated frame.

quant-ph↗

arXiv2012

Entanglement behavior of quantum states of fermionic system in accelerated frame

In this article we investigate the entanglement behavior of quantum states of fermionic system in accelerated frame. It was known that unlike scalar case the entanglement of fermionic maximally entangled states survives even in the infinite acceleration limit. Also for the fermionic system it was discussed that one may consider the approach beyond single mode approximation. However due to the peculiar property of fermionic system there has been different opinions about correct physical structure of fermionic system. Recently Montero and Martín-Martínez suggested an approach beyond single mode approximation. Using the structure proposed by Montero and Martín-Martínez we investigate the entanglement behavior of quantum states of fermionic system in accelerated frame. We find that the approach by Montero and Martín-Martínez seems to work appropriately. Also we may obtain the proper entanglement behavior of quantum states in fermionic system in accelerated frame(our result may correct the previous results published in Phys.Rev.A 83 052306).

quant-ph↗

arXiv2012

Entanglement amplification of fermionic systems in an accelerated frame

In this article we present an analysis to derive physical results in the entanglement amplification of fermonic systems in the relativistic regime, that is, beyond the single-mode approximation. This leads a recent work in [M. Montero and E. Martín-Martínez, JEHP 07 (2011) 006] to a physical result, and solidifies that phenomenon of entanglement amplification can actually happen in the relativistic regime.

quant-ph↗