arXiv subjects

Chu-Song Chen

Publications and source records attributed to Chu-Song Chen.

At least 19 recordsLinked to original sources

arXiv2026

Multimodal Graph RAG for Long-range Visually Rich Document Understanding

Multimodal large language models (MLLMs) are widely applied to visual document understanding. However, comprehending long documents remains an issue by the limited context window. Though recent multimodal retrieval-augmented generation (MMRAG) can address this challenge by retrieving relevant pages. It still struggles with the visual question answering (VQA) requiring holistic comprehension of a document. To cope with this, knowledge graph (KG) that summarizes global knowledge of a document can provide an effective solution. However, most existing LLM-based KG construction methods handle only the language modality, leaving the automatic creation of multimodal KGs (MMKGs) for visually rich documents largely unexplored. In this paper, we introduce a multimodal graph-based RAG approach to tackle this problem. Existing LLM-based KG methods evaluate the QA performance relying on indirect evidence such as comprehensiveness, diversity, empowerment, and so on. The lack of annotated datasets for comprehensive document-level VQA poses a significant challenge to effective model evaluation. To overcome this limitation, we also introduce a new benchmark, DLVQA (document-level VQA), which provides reference summaries and corresponding supporting facts for global document-level questions. Experimental results show that our approach outperforms existing MMRAG or KG-based approaches on multi-hop QA/VQA benchmarks and DLVQA.

cs.IR↗

arXiv2026

Document-Level Numerical Reasoning across Single and Multiple Tables in Financial Reports

Despite the strong language understanding abilities of large language models (LLMs), they still struggle with reliable question answering (QA) over long, structured documents, particularly for numerical reasoning. Financial annual reports exemplify this difficulty: financial statement analysis often hinges on accurate arithmetic, and analysts derive key indicators by integrating evidence scattered across multiple tables and narrative text. However, existing benchmarks focus largely on single-table settings, leaving cross-table document-level numerical reasoning underexplored. To address this gap, we introduce FinLongDocQA, a dataset for both single-table and cross-table financial numerical reasoning in long-context reports. Evaluating both closed-source and open-source LLMs on FinLongDocQA reveals two bottlenecks: (1) annual reports often exceed 129k tokens, exacerbating the context rot problem for locating relevant tables; and (2) even when relevant evidence is located, LLMs remain prone to errors in multi-step numerical reasoning. We propose FinLongDocAgent, a Multi-Agent Multi-Round Retrieval-Augmented Generation (RAG) approach that iteratively retrieves evidence, performs intermediate calculations, and verifies results across rounds. Experiments highlight the importance of iterative retrieval and verification for reliable numerical QA in long financial documents.

cs.CL↗

arXiv2026

Customized Visual Storytelling with Unified Multimodal LLMs

Multimodal story customization aims to generate coherent story flows conditioned on textual descriptions, reference identity images, and shot types. While recent progress in story generation has shown promising results, most approaches rely on text-only inputs. A few studies incorporate character identity cues (e.g., facial ID), but lack broader multimodal conditioning. In this work, we introduce VstoryGen, a multimodal framework that integrates descriptions with character and background references to enable customizable story generation. To enhance cinematic diversity, we introduce shot-type control via parameter-efficient prompt tuning on movie data, enabling the model to generate sequences that more faithfully reflect cinematic grammar. To evaluate our framework, we establish two new benchmarks that assess multimodal story customization from the perspectives of character and scene consistency, text-visual alignment, and shot-type control. Experiments demonstrate that VstoryGen achieves improved consistency and cinematic diversity compared to existing methods.

cs.CV↗

arXiv2025

MegaRAG: Multimodal Knowledge Graph-Based Retrieval Augmented Generation

Retrieval-augmented generation (RAG) enables large language models (LLMs) to dynamically access external information, which is powerful for answering questions over previously unseen documents. Nonetheless, they struggle with high-level conceptual understanding and holistic comprehension due to limited context windows, which constrain their ability to perform deep reasoning over long-form, domain-specific content such as full-length books. To solve this problem, knowledge graphs (KGs) have been leveraged to provide entity-centric structure and hierarchical summaries, offering more structured support for reasoning. However, existing KG-based RAG solutions remain restricted to text-only inputs and fail to leverage the complementary insights provided by other modalities such as vision. On the other hand, reasoning from visual documents requires textual, visual, and spatial cues into structured, hierarchical concepts. To address this issue, we introduce a multimodal knowledge graph-based RAG that enables cross-modal reasoning for better content understanding. Our method incorporates visual cues into the construction of knowledge graphs, the retrieval phase, and the answer generation process. Experimental results across both global and fine-grained question answering tasks show that our approach consistently outperforms existing RAG-based approaches on both textual and multimodal corpora.

cs.AI↗

arXiv2025

Mask-aware Text-to-Image Retrieval: Referring Expression Segmentation Meets Cross-modal Retrieval

Text-to-image retrieval (TIR) aims to find relevant images based on a textual query, but existing approaches are primarily based on whole-image captions and lack interpretability. Meanwhile, referring expression segmentation (RES) enables precise object localization based on natural language descriptions but is computationally expensive when applied across large image collections. To bridge this gap, we introduce Mask-aware TIR (MaTIR), a new task that unifies TIR and RES, requiring both efficient image search and accurate object segmentation. To address this task, we propose a two-stage framework, comprising a first stage for segmentation-aware image retrieval and a second stage for reranking and object grounding with a multimodal large language model (MLLM). We leverage SAM 2 to generate object masks and Alpha-CLIP to extract region-level embeddings offline at first, enabling effective and scalable online retrieval. Secondly, MLLM is used to refine retrieval rankings and generate bounding boxes, which are matched to segmentation masks. We evaluate our approach on COCO and D$^3$ datasets, demonstrating significant improvements in both retrieval accuracy and segmentation quality over previous methods.

cs.CV↗

arXiv2025

Relation-Rich Visual Document Generator for Visual Information Extraction

Despite advances in Large Language Models (LLMs) and Multimodal LLMs (MLLMs) for visual document understanding (VDU), visual information extraction (VIE) from relation-rich documents remains challenging due to the layout diversity and limited training data. While existing synthetic document generators attempt to address data scarcity, they either rely on manually designed layouts and templates, or adopt rule-based approaches that limit layout diversity. Besides, current layout generation methods focus solely on topological patterns without considering textual content, making them impractical for generating documents with complex associations between the contents and layouts. In this paper, we propose a Relation-rIch visual Document GEnerator (RIDGE) that addresses these limitations through a two-stage approach: (1) Content Generation, which leverages LLMs to generate document content using a carefully designed Hierarchical Structure Text format which captures entity categories and relationships, and (2) Content-driven Layout Generation, which learns to create diverse, plausible document layouts solely from easily available Optical Character Recognition (OCR) results, requiring no human labeling or annotations efforts. Experimental results have demonstrated that our method significantly enhances the performance of document understanding models on various VIE benchmarks. The code and model will be available at https://github.com/AI-Application-and-Integration-Lab/RIDGE .

cs.CV↗

arXiv2025

Safety Alignment Depth in Large Language Models: A Markov Chain Perspective

Large Language Models (LLMs) are increasingly adopted in high-stakes scenarios, yet their safety mechanisms often remain fragile. Simple jailbreak prompts or even benign fine-tuning can bypass these protocols, underscoring the need to understand where and how they fail. Recent findings suggest that vulnerabilities emerge when alignment is confined to only the initial output tokens. Unfortunately, even with the introduction of deep safety alignment, determining the optimal safety depth remains an unresolved challenge. By leveraging the equivalence between autoregressive language models and Markov chains, this paper offers the first theoretical result on how to identify the ideal depth for safety alignment, and demonstrates how permutation-based data augmentation can tighten these bounds. Crucially, we reveal a fundamental interaction between alignment depth and ensemble width-indicating that broader ensembles can compensate for shallower alignments. These insights provide a theoretical foundation for designing more robust, scalable safety strategies that complement existing alignment approaches, opening new avenues for research into safer, more reliable LLMs.

cs.LG↗

arXiv2024

Open-Vocabulary Panoptic Segmentation Using BERT Pre-Training of Vision-Language Multiway Transformer Model

Open-vocabulary panoptic segmentation remains a challenging problem. One of the biggest difficulties lies in training models to generalize to an unlimited number of classes using limited categorized training data. Recent popular methods involve large-scale vision-language pre-trained foundation models, such as CLIP. In this paper, we propose OMTSeg for open-vocabulary segmentation using another large-scale vision-language pre-trained model called BEiT-3 and leveraging the cross-modal attention between visual and linguistic features in BEiT-3 to achieve better performance. Experiments result demonstrates that OMTSeg performs favorably against state-of-the-art models.

cs.CV↗

arXiv2024

ACCEPT: Adaptive Codebook for Composite and Efficient Prompt Tuning

Prompt Tuning has been a popular Parameter-Efficient Fine-Tuning method attributed to its remarkable performance with few updated parameters on various large-scale pretrained Language Models (PLMs). Traditionally, each prompt has been considered indivisible and updated independently, leading the parameters increase proportionally as prompt length grows. To address this issue, we propose Adaptive Codebook for Composite and Efficient Prompt Tuning (ACCEPT). In our method, we refer to the concept of product quantization (PQ), allowing all soft prompts to share a set of learnable codebook vectors in each subspace, with each prompt differentiated by a set of adaptive weights. We achieve the superior performance on 17 diverse natural language tasks including natural language understanding (NLU) and question answering (QA) tasks by tuning only 0.3% of parameters of the PLMs. Our approach also excels in few-shot and large model settings, highlighting its significant potential.

cs.CL↗

arXiv2024

The Great Contradiction Showdown: How Jailbreak and Stealth Wrestle in Vision-Language Models?

Vision-Language Models (VLMs) have achieved remarkable performance on a variety of tasks, yet they remain vulnerable to jailbreak attacks that compromise safety and reliability. In this paper, we provide an information-theoretic framework for understanding the fundamental trade-off between the effectiveness of these attacks and their stealthiness. Drawing on Fano's inequality, we demonstrate how an attacker's success probability is intrinsically linked to the stealthiness of generated prompts. Building on this, we propose an efficient algorithm for detecting non-stealthy jailbreak attacks, offering significant improvements in model robustness. Experimental results highlight the tension between strong attacks and their detectability, providing insights into both adversarial strategies and defense mechanisms.

cs.LG↗

arXiv2024

SAM4MLLM: Enhance Multi-Modal Large Language Model for Referring Expression Segmentation

We introduce SAM4MLLM, an innovative approach which integrates the Segment Anything Model (SAM) with Multi-Modal Large Language Models (MLLMs) for pixel-aware tasks. Our method enables MLLMs to learn pixel-level location information without requiring excessive modifications to the existing model architecture or adding specialized tokens. We introduce an inquiry-based approach that can effectively find prompt points for SAM to perform segmentation based on MLLM. It combines detailed visual information with the powerful expressive capabilities of large language models in a unified language-based manner without additional computational overhead in learning. Experimental results on pubic benchmarks demonstrate the effectiveness of our approach.

cs.AI↗

arXiv2024

Defending Against Repetitive Backdoor Attacks on Semi-supervised Learning through Lens of Rate-Distortion-Perception Trade-off

Semi-supervised learning (SSL) has achieved remarkable performance with a small fraction of labeled data by leveraging vast amounts of unlabeled data from the Internet. However, this large pool of untrusted data is extremely vulnerable to data poisoning, leading to potential backdoor attacks. Current backdoor defenses are not yet effective against such a vulnerability in SSL. In this study, we propose a novel method, Unlabeled Data Purification (UPure), to disrupt the association between trigger patterns and target classes by introducing perturbations in the frequency domain. By leveraging the Rate-Distortion-Perception (RDP) trade-off, we further identify the frequency band, where the perturbations are added, and justify this selection. Notably, UPure purifies poisoned unlabeled data without the need of extra clean labeled data. Extensive experiments on four benchmark datasets and five SSL algorithms demonstrate that UPure effectively reduces the attack success rate from 99.78% to 0% while maintaining model accuracy. Code is available here: \url{https://github.com/chengyi-chris/UPure}.

cs.CV↗

arXiv2024

RDPN6D: Residual-based Dense Point-wise Network for 6Dof Object Pose Estimation Based on RGB-D Images

In this work, we introduce a novel method for calculating the 6DoF pose of an object using a single RGB-D image. Unlike existing methods that either directly predict objects' poses or rely on sparse keypoints for pose recovery, our approach addresses this challenging task using dense correspondence, i.e., we regress the object coordinates for each visible pixel. Our method leverages existing object detection methods. We incorporate a re-projection mechanism to adjust the camera's intrinsic matrix to accommodate cropping in RGB-D images. Moreover, we transform the 3D object coordinates into a residual representation, which can effectively reduce the output space and yield superior performance. We conducted extensive experiments to validate the efficacy of our approach for 6D pose estimation. Our approach outperforms most previous methods, especially in occlusion scenarios, and demonstrates notable improvements over the state-of-the-art methods. Our code is available on https://github.com/AI-Application-and-Integration-Lab/RDPN6D.

cs.CV↗

arXiv2023

D4AM: A General Denoising Framework for Downstream Acoustic Models

The performance of acoustic models degrades notably in noisy environments. Speech enhancement (SE) can be used as a front-end strategy to aid automatic speech recognition (ASR) systems. However, existing training objectives of SE methods are not fully effective at integrating speech-text and noisy-clean paired data for training toward unseen ASR systems. In this study, we propose a general denoising framework, D4AM, for various downstream acoustic models. Our framework fine-tunes the SE model with the backward gradient according to a specific acoustic model and the corresponding classification objective. In addition, our method aims to consider the regression objective as an auxiliary loss to make the SE model generalize to other unseen acoustic models. To jointly train an SE unit with regression and classification objectives, D4AM uses an adjustment scheme to directly estimate suitable weighting coefficients rather than undergoing a grid search process with additional training costs. The adjustment scheme consists of two parts: gradient calibration and regression objective weighting. The experimental results show that D4AM can consistently and effectively provide improvements to various unseen acoustic models and outperforms other combination setups. Specifically, when evaluated on the Google ASR API with real noisy data completely unseen during SE training, D4AM achieves a relative WER reduction of 24.65% compared with the direct feeding of noisy input. To our knowledge, this is the first work that deploys an effective combination scheme of regression (denoising) and classification (ASR) objectives to derive a general pre-processor applicable to various unseen ASR systems. Our code is available at https://github.com/ChangLee0903/D4AM.

cs.SD↗

arXiv2023

LC4SV: A Denoising Framework Learning to Compensate for Unseen Speaker Verification Models

The performance of speaker verification (SV) models may drop dramatically in noisy environments. A speech enhancement (SE) module can be used as a front-end strategy. However, existing SE methods may fail to bring performance improvements to downstream SV systems due to artifacts in the predicted signals of SE models. To compensate for artifacts, we propose a generic denoising framework named LC4SV, which can serve as a pre-processor for various unknown downstream SV models. In LC4SV, we employ a learning-based interpolation agent to automatically generate the appropriate coefficients between the enhanced signal and its noisy input to improve SV performance in noisy environments. Our experimental results demonstrate that LC4SV consistently improves the performance of various unseen SV systems. To the best of our knowledge, this work is the first attempt to develop a learning-based interpolation scheme aiming at improving SV performance in noisy environments.

eess.AS↗

arXiv2023

Domain-Generalized Face Anti-Spoofing with Unknown Attacks

Although face anti-spoofing (FAS) methods have achieved remarkable performance on specific domains or attack types, few studies have focused on the simultaneous presence of domain changes and unknown attacks, which is closer to real application scenarios. To handle domain-generalized unknown attacks, we introduce a new method, DGUA-FAS, which consists of a Transformer-based feature extractor and a synthetic unknown attack sample generator (SUASG). The SUASG network simulates unknown attack samples to assist the training of the feature extractor. Experimental results show that our method achieves superior performance on domain generalization FAS with known or unknown attacks.

cs.CV↗

arXiv2022

Globally Consistent Video Depth and Pose Estimation with Efficient Test-Time Training

Dense depth and pose estimation is a vital prerequisite for various video applications. Traditional solutions suffer from the robustness of sparse feature tracking and insufficient camera baselines in videos. Therefore, recent methods utilize learning-based optical flow and depth prior to estimate dense depth. However, previous works require heavy computation time or yield sub-optimal depth results. We present GCVD, a globally consistent method for learning-based video structure from motion (SfM) in this paper. GCVD integrates a compact pose graph into the CNN-based optimization to achieve globally consistent estimation from an effective keyframe selection mechanism. It can improve the robustness of learning-based methods with flow-guided keyframes and well-established depth prior. Experimental results show that GCVD outperforms the state-of-the-art methods on both depth and pose estimation. Besides, the runtime experiments reveal that it provides strong efficiency in both short- and long-term videos with global consistency provided.

cs.CV↗

arXiv2022

NASTAR: Noise Adaptive Speech Enhancement with Target-Conditional Resampling

For deep learning-based speech enhancement (SE) systems, the training-test acoustic mismatch can cause notable performance degradation. To address the mismatch issue, numerous noise adaptation strategies have been derived. In this paper, we propose a novel method, called noise adaptive speech enhancement with target-conditional resampling (NASTAR), which reduces mismatches with only one sample (one-shot) of noisy speech in the target environment. NASTAR uses a feedback mechanism to simulate adaptive training data via a noise extractor and a retrieval model. The noise extractor estimates the target noise from the noisy speech, called pseudo-noise. The noise retrieval model retrieves relevant noise samples from a pool of noise signals according to the noisy speech, called relevant-cohort. The pseudo-noise and the relevant-cohort set are jointly sampled and mixed with the source speech corpus to prepare simulated training data for noise adaptation. Experimental results show that NASTAR can effectively use one noisy speech sample to adapt an SE model to a target condition. Moreover, both the noise extractor and the noise retrieval model contribute to model adaptation. To our best knowledge, NASTAR is the first work to perform one-shot noise adaptation through noise extraction and retrieval.

eess.AS↗