Seongjae Hwang

Sujung Hong, Chanyong Yoon, Seongjae Hwang

Large diffusion vision-language models (LDVLMs) have recently emerged as a promising alternative to autoregressive models, enabling parallel decoding for efficient inference and leveraging bidirectional attention for global context. Despite these advances, their behavior under long-form generation remains underexplored. In this work, we show that existing LDVLMs suffer from repetitive generation and degraded visual grounding, and identify two underlying causes. First, repetitive generation originates from a mask token prior: since generation tokens are initialized as mask tokens, their hidden representations progressively drift toward a shared prior direction over generation steps. Second, a fundamental misalignment between the positional attention bias and the iterative unmasking process suppresses attention toward informative visual tokens, degrading visual grounding. Based on these insights, we propose a training-free approach, introducing Mask Prior Suppression and Monotonic RoPE Scaling to mitigate mask prior drift and positional attention collapse during decoding. Experiments on general multimodal benchmarks and visual grounding tasks demonstrate improvements over baseline LDVLMs, with robust gains on long-form description benchmarks. Our results show that these failures can be effectively addressed with a lightweight, plug-and-play strategy that requires no additional training and generalizes across diverse LDVLM architectures.

Jaehoon Joo, Taejin Jeong, Seongjae Hwang

Understanding how humans process visual information is one of the crucial steps for unraveling the underlying mechanism of brain activity. Recently, this curiosity has motivated the fMRI-to-image reconstruction task; given the fMRI data from visual stimuli, it aims to reconstruct the corresponding visual stimuli. Surprisingly, leveraging powerful generative models such as the Latent Diffusion Model (LDM) has shown promising results in reconstructing complex visual stimuli such as high-resolution natural images from vision datasets. Despite the impressive structural fidelity of these reconstructions, they often lack details of small objects, ambiguous shapes, and semantic nuances. Consequently, the incorporation of additional semantic knowledge, beyond mere visuals, becomes imperative. In light of this, we exploit how modern LDMs effectively incorporate multi-modal guidance (text guidance, visual guidance, and image layout) for structurally and semantically plausible image generations. Specifically, inspired by the two-streams hypothesis suggesting that perceptual and semantic information are processed in different brain regions, our framework, Brain-Streams, maps fMRI signals from these brain regions to appropriate embeddings. That is, by extracting textual guidance from semantic information regions and visual guidance from perceptual information regions, Brain-Streams provides accurate multi-modal guidance to LDMs. We validate the reconstruction ability of Brain-Streams both quantitatively and qualitatively on a real fMRI dataset comprising natural image stimuli and fMRI data.

Won Hwa Kim, Mona Jalal, Seongjae Hwang et al.

The adoption of "human-in-the-loop" paradigms in computer vision and machine learning is leading to various applications where the actual data acquisition (e.g., human supervision) and the underlying inference algorithms are closely interwined. While classical work in active learning provides effective solutions when the learning module involves classification and regression tasks, many practical issues such as partially observed measurements, financial constraints and even additional distributional or structural aspects of the data typically fall outside the scope of this treatment. For instance, with sequential acquisition of partial measurements of data that manifest as a matrix (or tensor), novel strategies for completion (or collaborative filtering) of the remaining entries have only been studied recently. Motivated by vision problems where we seek to annotate a large dataset of images via a crowdsourced platform or alternatively, complement results from a state-of-the-art object detector using human feedback, we study the "completion" problem defined on graphs, where requests for additional measurements must be made sequentially. We design the optimization model in the Fourier domain of the graph describing how ideas based on adaptive submodularity provide algorithms that work well in practice. On a large set of images collected from Imgur, we see promising results on images that are otherwise difficult to categorize. We also show applications to an experimental design problem in neuroimaging.