arXiv subjects

Yuexian Zou

Publications and source records attributed to Yuexian Zou.

At least 19 recordsLinked to original sources

arXiv2026

Persistent Delivery Optimization for Streaming Speech-to-Text Translation with Revisions

Revision-capable streaming speech-to-text translation (S2TT) can correct earlier drafts, but process rewards based on visible text may credit content later withdrawn. Persistent Delivery Optimization (PDO) assigns intermediate reward only to content that survives revisions while scoring final quality separately. With 7.49 h of task-specific FLEURS adaptation, PDO achieves the best BLEU on four of five directions and higher COMET than every external streaming baseline in all five directions. Relative to its History-SFT initialization, PDO reduces mean/P90 finalization-aware latency by 10.8\%/11.3\% and normalized erasure by 15.8\%, while emitting at the first permitted 2-s update and improving macro BLEU. Zero-shot evaluation on Europarl-ST and CoVoST 2 confirms that these gains are not confined to the FLEURS training domain.

eess.AS↗

arXiv2026

Are Prompt Optimizers Blind? Cross-Modal Visual Feedback for Automatic Prompt Optimization

Automatic prompt optimization (APO) has been widely adopted to adapt vision-language models (VLMs) to downstream tasks without weight updates, yielding promising results. However, on multimodal tasks, the effectiveness of APO is fundamentally bottlenecked by a blind feedback channel: the optimizer reads the question, the prediction, and the gold answer, but never the input image on which the model failed, and therefore cannot diagnose visually grounded errors. As a remedy, we introduce Cross-Modal Visual Feedback (CMVF). CMVF incorporates (1) a failure-conditioned visual diagnosis stage, in which a stronger optimizer VLM inspects each failed image without access to predictions or labels, and (2) an error-aware aggregation stage that compresses these observations into reusable, task-level visual blind-spot patterns that drive the prompt rewrite. Crucially, the image is consumed only during optimization; the deployed artifact is an ordinary text prompt that runs at the same inference cost as any text-only baseline. Extensive results across 12 VQA datasets and 4 target VLMs demonstrate that CMVF consistently ranks first, improving over the strongest baseline on every target by 2.4 points on average, with gains of up to 6.5 points on individual benchmarks. Moreover, the optimizer self-organizes into expert-style visual checklists that transfer across models without re-optimization.

cs.AI↗

arXiv2026

HeartMuLa: A Family of Open Sourced Music Foundation Models

We present a family of open-source Music Foundation Models designed to advance large-scale music understanding and generation across diverse tasks and modalities. Our framework consists of four major components: (1) HeartCLAP, an audio-text alignment model; (2) HeartTranscriptor, a robust lyric recognition model optimized for real-world music scenarios; and (3) HeartCodec, a low-frame-rate (12.5 Hz) yet high-fidelity music codec tokenizer that captures long-range musical structure while preserving fine-grained acoustic details and enabling efficient autoregressive modeling; (4) HeartMuLa, an LLM-based song generation model capable of synthesizing high-fidelity music under rich, user-controllable conditions (e.g., textual style descriptions, lyrics, and reference audio). In addition, it provides two specialized modes: (i) fine-grained musical attribute control, which allows users to specify the style of different song sections (e.g., intro, verse, chorus) using natural language prompts; and (ii) short, engaging music generation, which is suitable as background music for short videos. Lastly, HeartMuLa improves significantly when scaled to 7B parameters. For the first time, we show that a Suno-level, commercial-grade system can be reproduced using academic-scale data and GPU resources. We expect these foundation models to serve as strong baselines for future research and to facilitate practical applications in multimodal content production.

cs.SD↗

arXiv2026

Graph-PiT: Enhancing Structural Coherence in Part-Based Image Synthesis via Graph Priors

Achieving fine-grained and structurally sound controllability is a cornerstone of advanced visual generation. Existing part-based frameworks treat user-provided parts as an unordered set and therefore ignore their intrinsic spatial and semantic relationships, which often results in compositions that lack structural integrity. To bridge this gap, we propose Graph-PiT, a framework that explicitly models the structural dependencies of visual components using a graph prior. Specifically, we represent visual parts as nodes and their spatial-semantic relationships as edges. At the heart of our method is a Hierarchical Graph Neural Network (HGNN) module that performs bidirectional message passing between coarse-grained part-level super-nodes and fine-grained IP+ token sub-nodes, refining part embeddings before they enter the generative pipeline. We also introduce a graph Laplacian smoothness loss and an edge-reconstruction loss so that adjacent parts acquire compatible, relation-aware embeddings. Quantitative experiments on controlled synthetic domains (character, product, indoor layout, and jigsaw), together with qualitative transfer to real web images, show that Graph-PiT improves structural coherence over vanilla PiT while remaining compatible with the original IP-Prior pipeline. Ablation experiments confirm that explicit relational reasoning is crucial for enforcing user-specified adjacency constraints. Our approach not only enhances the plausibility of generated concepts but also offers a scalable and interpretable mechanism for complex, multi-part image synthesis. The code is available at https://github.com/wolf-bailang/Graph-PiT.

cs.CV↗

arXiv2026

SemanticVocoder: Bridging Audio Generation and Audio Understanding via Semantic Latents

Recent audio generation models typically rely on Variational Autoencoders (VAEs) and perform generation within the VAE latent space. Although VAEs excel at compression and reconstruction, their latents inherently encode low-level acoustic details rather than semantically discriminative information, leading to entangled event semantics and complicating the training of generative models. To address these issues, we discard VAE acoustic latents and introduce semantic encoder latents, thereby proposing SemanticVocoder, a generative vocoder that directly synthesizes waveforms from semantic latents. Equipped with SemanticVocoder, our text-to-audio generation model achieves a Frechet Distance of 12.823 and a Frechet Audio Distance of 1.709 on the AudioCaps test set, as the introduced semantic latents exhibit superior discriminability compared to acoustic VAE latents. Beyond improved generation performance, it also serves as a promising attempt towards unifying audio understanding and generation within a shared semantic space. Generated samples are available at https://zeyuxie29.github.io/SemanticVocoder/.

cs.SD↗

arXiv2026

PhoenixCodec: Taming Neural Speech Coding for Extreme Low-Resource Scenarios

This paper presents PhoenixCodec, a comprehensive neural speech coding and decoding framework designed for extremely low-resource conditions. The proposed system integrates an optimized asymmetric frequency-time architecture, a Cyclical Calibration and Refinement (CCR) training strategy, and a noise-invariant fine-tuning procedure. Under stringent constraints - computation below 700 MFLOPs, latency less than 30 ms, and dual-rate support at 1 kbps and 6 kbps - existing methods face a trade-off between efficiency and quality. PhoenixCodec addresses these challenges by alleviating the resource scattering of conventional decoders, employing CCR to enhance optimization stability, and enhancing robustness through noisy-sample fine-tuning. In the LRAC 2025 Challenge Track 1, the proposed system ranked third overall and demonstrated the best performance at 1 kbps in both real-world noise and reverberation and intelligibility in clean tests, confirming its effectiveness.

eess.AS↗

arXiv2026

Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning

Recent work on test-time scaling for large language model (LLM) reasoning typically assumes that allocating more inference-time computation uniformly improves correctness. However, prior studies show that reasoning uncertainty is highly localized: a small subset of low-confidence tokens disproportionately contributes to reasoning errors and unnecessary output expansion. Motivated by this observation, we propose Thinking by Subtraction, a confidence-driven contrastive decoding approach that improves reasoning reliability through targeted token-level intervention. Our method, Confidence-Driven Contrastive Decoding, detects low-confidence tokens during decoding and intervenes selectively at these positions. It constructs a contrastive reference by replacing high-confidence tokens with minimal placeholders, and refines predictions by subtracting this reference distribution at low-confidence locations. Experiments show that CCD significantly improves accuracy across mathematical reasoning benchmarks while substantially reducing output length, with minimal KV-cache overhead. As a training-free method, CCD enhances reasoning reliability through targeted low-confidence intervention without computational redundancy. Our code will be made available at: https://github.com/bolo-web/CCD.

cs.CL↗

arXiv2026

DermoGPT: Open Weights and Open Data for Morphology-Grounded Dermatological Reasoning MLLMs

Multimodal Large Language Models (MLLMs) show promise for medical applications, yet progress in dermatology lags due to limited training data, narrow task coverage, and lack of clinically-grounded supervision that mirrors expert diagnostic workflows. We present a comprehensive framework to address these gaps. First, we introduce DermoInstruct, a large-scale morphology-anchored instruction corpus comprising 211,243 images and 772,675 trajectories across five task formats, capturing the complete diagnostic pipeline from morphological observation and clinical reasoning to final diagnosis. Second, we establish DermoBench, a rigorous benchmark evaluating 11 tasks across four clinical axes: Morphology, Diagnosis, Reasoning, and Fairness, including a challenging subset of 3,600 expert-verified open-ended instances and human performance baselines. Third, we develop DermoGPT, a dermatology reasoning MLLM trained via supervised fine-tuning followed by our Morphologically-Anchored Visual-Inference-Consistent (MAVIC) reinforcement learning objective, which enforces consistency between visual observations and diagnostic conclusions. At inference, we deploy Confidence-Consistency Test-time adaptation (CCT) for robust predictions. Experiments show DermoGPT significantly outperforms 16 representative baselines across all axes, achieving state-of-the-art performance while substantially narrowing the human-AI gap. DermoInstruct, DermoBench and DermoGPT will be made publicly available at https://github.com/mendicant04/DermoGPT upon acceptance.

cs.CL↗

arXiv2025

MolSculpt: Sculpting 3D Molecular Geometries from Chemical Syntax

Generating precise 3D molecular geometries is crucial for drug discovery and material science. While prior efforts leverage 1D representations like SELFIES to ensure molecular validity, they fail to fully exploit the rich chemical knowledge entangled within 1D models, leading to a disconnect between 1D syntactic generation and 3D geometric realization. To bridge this gap, we propose MolSculpt, a novel framework that "sculpts" 3D molecular geometries from chemical syntax. MolSculpt is built upon a frozen 1D molecular foundation model and a 3D molecular diffusion model. We introduce a set of learnable queries to extract inherent chemical knowledge from the foundation model, and a trainable projector then injects this cross-modal information into the conditioning space of the diffusion model to guide the 3D geometry generation. In this way, our model deeply integrates 1D latent chemical knowledge into the 3D generation process through end-to-end optimization. Experiments demonstrate that MolSculpt achieves state-of-the-art (SOTA) performance in \textit{de novo} 3D molecule generation and conditional 3D molecule generation, showing superior 3D fidelity and stability on both the GEOM-DRUGS and QM9 datasets. Code is available at https://github.com/SakuraTroyChen/MolSculpt.

cs.LG↗

arXiv2025

U-Codec: Ultra Low Frame-rate Neural Speech Codec for Fast High-fidelity Speech Generation

We propose \textbf{U-Codec}, an \textbf{U}ltra low frame-rate neural speech \textbf{Codec} that achieves high-fidelity reconstruction and fast speech generation at an extremely low frame-rate of 5Hz (5 frames per second). Extreme compression at 5Hz typically leads to severe intelligibility and spectral detail loss, we introduce a Transformer-based inter-frame long-term dependency module and systematically explore residual vector quantization (RVQ) depth and codebook size to identify optimal configurations. Moreover, we apply U-Codec into a large language model (LLM)-based auto-regressive TTS model, which leverages global and local hierarchical architecture to effectively capture dependencies across multi-layer tokens. We extend LLM-based TTS from 3-layer RVQ at 50Hz to 32-layer RVQ at 5Hz. Experimental results demonstrate that U-Codec improves LLM-based TTS inference speed by around 3 $\times$ over high-frame-rate codecs while maintaining similarity and naturalness. These results validate the feasibility of using highly compressed 5Hz discrete tokens for fast and high-fidelity speech synthesis.

cs.SD↗

arXiv2025

BlobCtrl: Taming Controllable Blob for Element-level Image Editing

As user expectations for image editing continue to rise, the demand for flexible, fine-grained manipulation of specific visual elements presents a challenge for current diffusion-based methods. In this work, we present BlobCtrl, a framework for element-level image editing based on a probabilistic blob-based representation. Treating blobs as visual primitives, BlobCtrl disentangles layout from appearance, affording fine-grained, controllable object-level manipulation. Our key contributions are twofold: (1) an in-context dual-branch diffusion model that separates foreground and background processing, incorporating blob representations to explicitly decouple layout and appearance, and (2) a self-supervised disentangle-then-reconstruct training paradigm with an identity-preserving loss function, along with tailored strategies to efficiently leverage blob-image pairs. To foster further research, we introduce BlobData for large-scale training and BlobBench, a benchmark for systematic evaluation. Experimental results demonstrate that BlobCtrl achieves state-of-the-art performance in a variety of element-level editing tasks, such as object addition, removal, scaling, and replacement, while maintaining computational efficiency. Project Webpage: https://liyaowei-stu.github.io/project/BlobCtrl/

cs.CV↗

arXiv2025

IC-Custom: Diverse Image Customization via In-Context Learning

Image customization, a crucial technique for industrial media production, aims to generate content that is consistent with reference images. However, current approaches conventionally separate image customization into position-aware and position-free customization paradigms and lack a universal framework for diverse customization, limiting their applications across various scenarios. To overcome these limitations, we propose IC-Custom, a unified framework that seamlessly integrates position-aware and position-free image customization through in-context learning. IC-Custom concatenates reference images with target images to a polyptych, leveraging DiT's multi-modal attention mechanism for fine-grained token-level interactions. We propose the In-context Multi-Modal Attention (ICMA) mechanism, which employs learnable task-oriented register tokens and boundary-aware positional embeddings to enable the model to effectively handle diverse tasks and distinguish between inputs in polyptych configurations. To address the data gap, we curated a 12K identity-consistent dataset with 8K real-world and 4K high-quality synthetic samples, avoiding the overly glossy, oversaturated look typical of synthetic data. IC-Custom supports various industrial applications, including try-on, image insertion, and creative IP customization. Extensive evaluations on our proposed ProductBench and the publicly available DreamBench demonstrate that IC-Custom significantly outperforms community workflows, closed-source models, and state-of-the-art open-source approaches. IC-Custom achieves about 73\% higher human preference across identity consistency, harmony, and text alignment metrics, while training only 0.4\% of the original model parameters. Project page: https://liyaowei-stu.github.io/project/IC_Custom

cs.CV↗

arXiv2025

When Audio Generators Become Good Listeners: Generative Features for Understanding Tasks

This work pioneers the utilization of generative features in enhancing audio understanding. Unlike conventional discriminative features that directly optimize posterior and thus emphasize semantic abstraction while losing fine grained details, audio generation models inherently encode both spatiotemporal perception (capturing local acoustic texture across time and frequency) and semantic prior (knowing what to generate). It motivates us to explore the bridge of these complementary strengths. We provide a systematic investigation of their differences and complementary relationships, and ultimately propose an effective fusion strategy. Experiments across multiple tasks, including sound event classification, tagging, and particularly the fine grained task of audio captioning, demonstrate consistent performance gains. Beyond empirical improvements, this work more importantly introduces a new perspective on audio representation learning, highlighting that generative discriminative complementarity can provide both detailed perception and semantic awareness for audio understanding.

cs.SD↗

arXiv2025

FakeSound2: A Benchmark for Explainable and Generalizable Deepfake Sound Detection

The rapid development of generative audio raises ethical and security concerns stemming from forged data, making deepfake sound detection an important safeguard against the malicious use of such technologies. Although prior studies have explored this task, existing methods largely focus on binary classification and fall short in explaining how manipulations occur, tracing where the sources originated, or generalizing to unseen sources-thereby limiting the explainability and reliability of detection. To address these limitations, we present FakeSound2, a benchmark designed to advance deepfake sound detection beyond binary accuracy. FakeSound2 evaluates models across three dimensions: localization, traceability, and generalization, covering 6 manipulation types and 12 diverse sources. Experimental results show that although current systems achieve high classification accuracy, they struggle to recognize forged pattern distributions and provide reliable explanations. By highlighting these gaps, FakeSound2 establishes a comprehensive benchmark that reveals key challenges and aims to foster robust, explainable, and generalizable approaches for trustworthy audio authentication.

cs.SD↗

arXiv2025

STAR: Speech-to-Audio Generation via Representation Learning

This work presents STAR, the first end-to-end speech-to-audio generation framework, designed to enhance efficiency and address error propagation inherent in cascaded systems. Unlike prior approaches relying on text or vision, STAR leverages speech as it constitutes a natural modality for interaction. As an initial step to validate the feasibility of the system, we demonstrate through representation learning experiments that spoken sound event semantics can be effectively extracted from raw speech, capturing both auditory events and scene cues. Leveraging the semantic representations, STAR incorporates a bridge network for representation mapping and a two-stage training strategy to achieve end-to-end synthesis. With a 76.9% reduction in speech processing latency, STAR demonstrates superior generation performance over the cascaded systems. Overall, STAR establishes speech as a direct interaction signal for audio generation, thereby bridging representation learning and multimodal synthesis. Generated samples are available at https://zeyuxie29.github.io/STAR.

cs.SD↗

arXiv2025

Not All Tokens and Heads Are Equally Important: Dual-Level Attention Intervention for Hallucination Mitigation

Large vision-language models (LVLMs) have demonstrated impressive capabilities across diverse multimodal tasks, yet they remain highly susceptible to visual hallucinations (VH), often producing confident but inaccurate descriptions of visual content. Building on the insight that not all tokens and attention heads contribute equally to VH mitigation, we introduce VisFlow, a lightweight and training-free framework that alleviates hallucinations by directly modulating attention patterns during inference. To address two primary challenges of VH, namely insufficient visual attention and the dominance of language priors, we identify three problematic attention behaviors in LVLMs: (1) disproportionate allocation of attention to uninformative or trailing visual tokens, (2) over-dependence on the previously generated token, and (3) excessive fixation on system prompts that hinders multimodal integration. To overcome these issues, VisFlow introduces a dual-level Attention Intervention, consisting of Token-level Attention Intervention (TAI), which reinforces attention to salient visual regions, and Head-level Attention Intervention (HAI), which suppresses undue focus on system prompts and adjacent text tokens. Together, these interventions strengthen visual alignment while reducing linguistic bias. Extensive experiments across diverse models and benchmarks demonstrate that VisFlow effectively mitigates hallucinations with minimal computational overhead.

cs.CV↗

arXiv2025

ATRI: Mitigating Multilingual Audio Text Retrieval Inconsistencies by Reducing Data Distribution Errors

Multilingual audio-text retrieval (ML-ATR) is a challenging task that aims to retrieve audio clips or multilingual texts from databases. However, existing ML-ATR schemes suffer from inconsistencies for instance similarity matching across languages. We theoretically analyze the inconsistency in terms of both multilingual modal alignment direction error and weight error, and propose the theoretical weight error upper bound for quantifying the inconsistency. Based on the analysis of the weight error upper bound, we find that the inconsistency problem stems from the data distribution error caused by random sampling of languages. We propose a consistent ML-ATR scheme using 1-to-k contrastive learning and audio-English co-anchor contrastive learning, aiming to mitigate the negative impact of data distribution error on recall and consistency in ML-ATR. Experimental results on the translated AudioCaps and Clotho datasets show that our scheme achieves state-of-the-art performance on recall and consistency metrics for eight mainstream languages, including English. Our code will be available at https://github.com/ATRI-ACL/ATRI-ACL.

cs.SD↗

arXiv2025

Do we really have to filter out random noise in pre-training data for language models?

Web-scale pre-training datasets are the cornerstone of LLMs' success. However, text data curated from the Internet inevitably contains random noise caused by decoding errors or unregulated web content. In contrast to previous works that focus on low quality or synthetic data, our study \textbf{provides the first systematic investigation of such random noise through a cohesive ``What-Why-How'' framework.} Surprisingly, we observed that the resulting increase in the loss of next-token prediction (NTP) was significantly lower than the proportion of random noise even when the model was scaled up to 2.7B. We provide a theoretical justification for this phenomenon, which also elucidates the success of multilingual models and can be applied to multimodal models. On the other hand, experiments show that the model's performance in downstream tasks is not based solely on the NTP loss, which means that random noise may result in degraded downstream performance. To address the potential adverse effects, we introduce a novel plug-and-play Local Gradient Matching loss, which explicitly enhances the denoising capability of the downstream task head by aligning the gradient of normal and perturbed features without requiring knowledge of the model's parameters. Additional experiments on 8 language and 14 vision benchmarks further validate its effectiveness.

cs.CL↗