arXiv subjects

Xulong Zhang

Publications and source records attributed to Xulong Zhang.

At least 19 recordsLinked to original sources

arXiv2026

ProbPlug: A Plugin Uncertainty Network for Reliable Confidence in LLM Binary Classification

Large language models (LLMs) have achieved strong performance across a broad range of classification settings, yet the reliability of their predictions remains a major obstacle to deployment in high-stakes scenarios. Although confidence estimation for LLMs has been widely studied, confidence calibration for LLM-based classification remains underexplored. We introduce ProbPlug, a lightweight confidence estimation framework for LLM-based binary classification, which predicts whether an output is correct using internal token features extracted from a frozen LLM. ProbPlug employs a self-attention module to aggregate hidden representations and can be integrated into the original inference pipeline without modifying the base model. Experiments across multiple tasks involving both text-based and multimodal large models show that ProbPlug provides more reliable confidence estimates, improves classification performance with negligible additional overhead, and exhibits strong generalization across tasks. These results indicate that ProbPlug serves as a practical solution for confidence estimation in LLM-based classification. Our code is publicly available at Github.

cs.CL↗

arXiv2026

DIVA: Harnessing the Representation Divergence in Unified Multimodal Models for Mutual Reinforcement

Unified Multimodal models (UMMs) built on a single architecture have shown impressive performance in both understanding and generation. We identify a fundamental challenge that lies in inductive biases induced by distinct supervision signals: generation branch prefers high-fidelity, fine-grained representations capable of reconstruction, while the understanding favours semantically discriminative embeddings that remain invariant to task-irrelevant factors. Consequently, optimizing these complementary but non-equivalent objectives within a monolithic backbone leads to mutual impairment instead of enhancement. In this paper, we first analyze the root cause of this interference in unified backbones and reveal a complementary structure in their internal representations. Motivated by the observation, we propose DIVA, a self-improved post-training framework that transforms the representation divergence into interior synergy. By explicitly factorizing the visual representation into shared and unique components based on two complementary information flow, DIVA enables both the understanding and generation branches to achieve beneficial transferring while preserving the integrity of unique information from cross-flow interference via mutual information estimation. Despite its generality, our method consistently achieves improvements across visual understanding (+7.82%) and generation (+8.46%). The official code is available at: https://github.com/Jayyy-H/DIVA.

cs.CV↗

arXiv2026

Evolvable Embodied Agent for Robotic Manipulation via Long Short-Term Reflection and Optimization

Achieving general-purpose robotics requires empowering robots to adapt and evolve based on their environment and feedback. Traditional methods face limitations such as extensive training requirements, difficulties in cross-task generalization, and lack of interpretability. Prompt learning offers new opportunities for self-evolving robots without extensive training, but simply reflecting on past experiences. However, extracting meaningful insights from task successes and failures remains a challenge. To this end, we propose the evolvable embodied agent (EEAgent) framework, which leverages large vision-language models (VLMs) for better environmental interpretation and policy planning. To enhance reflection on past experiences, we propose a long short-term reflective optimization (LSTRO) mechanism that dynamically refines prompts based on both past experiences and newly learned lessons, facilitating continuous self-evolution, thereby enhancing overall task success rates. Evaluations on six VIMA-Bench tasks reveal that our approach sets a new state-of-the-art, notably outperforming baselines in complex scenarios.

cs.RO↗

arXiv2026

From Inheritance to Saturation: Disentangling the Evolution of Visual Redundancy for Architecture-Aware MLLM Inference Acceleration

High-resolution Multimodal Large Language Models (MLLMs) face prohibitive computational costs during inference due to the explosion of visual tokens. Existing acceleration strategies, such as token pruning or layer sparsity, suffer from severe "backbone dependency", performing well on Vicuna or Mistral architectures (e.g., LLaVA) but causing significant performance degradation when transferred to architectures like Qwen. To address this, we leverage truncated matrix entropy to uncover a universal three-stage inference lifecycle, decoupling visual redundancy into universal Intrinsic Visual Redundancy (IVR) and architecture-dependent Secondary Saturation Redundancy (SSR). Guided by this insight, we propose HalfV, a framework that first mitigates IVR via a unified pruning strategy and then adaptively handles SSR based on its specific manifestation. Experiments demonstrate that HalfV achieves superior efficiency-performance trade-offs across diverse backbones. Notably, on Qwen25-VL, it retains 96.8\% performance at a 4.1$\times$ FLOPs speedup, significantly outperforming state-of-the-art baselines. Our code is available at https://github.com/civilizwa/HalfV.

cs.CV↗

arXiv2026

Attention-weighted Centered Kernel Alignment for Knowledge Distillation in Large Audio-Language Models Applied to Speech Emotion Recognition

The emergence of Large Audio-Language Models (LALMs) has advanced Speech Emotion Recognition (SER), but their size limits deployment in resource-constrained environments. While Knowledge Distillation is effective for LALM compression, existing methods remain underexplored in distilling the cross-modal projection module (Projector), and often struggle with alignment due to differences in feature dimensions. We propose PL-Distill, a KD framework that combines Projector-Level Distillation (PDist) to align audio embeddings and Logits-Level Distillation (LDist) to align output logits. PDist introduces Attention-weighted Centered Kernel Alignment, a novel approach we propose to highlight important time steps and address dimension mismatches. Meanwhile, LDist minimizes the Kullback-Leibler divergence between teacher and student logits from audio and text modalities. On IEMOCAP, RAVDESS, and SAVEE, PL-Distill compresses an 8.4B-parameter teacher to a compact 1.1B-parameter student, consistently outperforming the teacher, state-of-the-art pretrained models, and other KD baselines across all metrics.

cs.SD↗

arXiv2026

CARE: Multi-Task Pretraining for Latent Continuous Action Representation in Robot Control

Recent advances in Vision-Language-Action (VLA) models have shown promise for robot control, but their dependence on action supervision limits scalability and generalization. To address this challenge, we introduce CARE, a novel framework designed to train VLA models for robotic task execution. Unlike existing methods that depend on action annotations during pretraining, CARE eliminates the need for explicit action labels by leveraging only video-text pairs. These weakly aligned data sources enable the model to learn continuous latent action representations through a newly designed multi-task pretraining objective. During fine-tuning, a small set of labeled data is used to train the action head for control. Experimental results across various simulation tasks demonstrate CARE's superior success rate, semantic interpretability, and ability to avoid shortcut learning. These results underscore CARE's scalability, interpretability, and effectiveness in robotic control with weak supervision.

cs.RO↗

arXiv2026

Head-Aware Visual Cropping: Enhancing Fine-Grained VQA with Attention-Guided Subimage

Multimodal Large Language Models (MLLMs) show strong performance in Visual Question Answering (VQA) but remain limited in fine-grained reasoning due to low-resolution inputs and noisy attention aggregation. We propose \textbf{Head Aware Visual Cropping (HAVC)}, a training-free method that improves visual grounding by leveraging a selectively refined subset of attention heads. HAVC first filters heads through an OCR-based diagnostic task, ensuring that only those with genuine grounding ability are retained. At inference, these heads are further refined using spatial entropy for stronger spatial concentration and gradient sensitivity for predictive contribution. The fused signals produce a reliable Visual Cropping Guidance Map, which highlights the most task-relevant region and guides the cropping of a subimage subsequently provided to the MLLM together with the image-question pair. Extensive experiments on multiple fine-grained VQA benchmarks demonstrate that HAVC consistently outperforms state-of-the-art cropping strategies, achieving more precise localization, stronger visual grounding, providing a simple yet effective strategy for enhancing precision in MLLMs.

cs.CV↗

arXiv2026

MIRRORTALK: Forging Personalized Avatars Via Disentangled Style and Hierarchical Motion Control

Synthesizing personalized talking faces that uphold and highlight a speaker's unique style while maintaining lip-sync accuracy remains a significant challenge. A primary limitation of existing approaches is the intrinsic confounding of speaker-specific talking style and semantic content within facial motions, which prevents the faithful transfer of a speaker's unique persona to arbitrary speech. In this paper, we propose MirrorTalk, a generative framework based on a conditional diffusion model, combined with a Semantically-Disentangled Style Encoder (SDSE) that can distill pure style representations from a brief reference video. To effectively utilize this representation, we further introduce a hierarchical modulation strategy within the diffusion process. This mechanism guides the synthesis by dynamically balancing the contributions of audio and style features across distinct facial regions, ensuring both precise lip-sync accuracy and expressive full-face dynamics. Extensive experiments demonstrate that MirrorTalk achieves significant improvements over state-of-the-art methods in terms of lip-sync accuracy and personalization preservation.

cs.CV↗

arXiv2025

Semi-Supervised Self-Learning Enhanced Music Emotion Recognition

Music emotion recognition (MER) aims to identify the emotions conveyed in a given musical piece. However, currently, in the field of MER, the available public datasets have limited sample sizes. Recently, segment-based methods for emotion-related tasks have been proposed, which train backbone networks on shorter segments instead of entire audio clips, thereby naturally augmenting training samples without requiring additional resources. Then, the predicted segment-level results are aggregated to obtain the entire song prediction. The most commonly used method is that the segment inherits the label of the clip containing it, but music emotion is not constant during the whole clip. Doing so will introduce label noise and make the training easy to overfit. To handle the noisy label issue, we propose a semi-supervised self-learning (SSSL) method, which can differentiate between samples with correct and incorrect labels in a self-learning manner, thus effectively utilizing the augmented segment-level data. Experiments on three public emotional datasets demonstrate that the proposed method can achieve better or comparable performance.

cs.SD↗

arXiv2025

CycleFlow: Leveraging Cycle Consistency in Flow Matching for Speaker Style Adaptation

Voice Conversion (VC) aims to convert the style of a source speaker, such as timbre and pitch, to the style of any target speaker while preserving the linguistic content. However, the ground truth of the converted speech does not exist in a non-parallel VC scenario, which induces the train-inference mismatch problem. Moreover, existing methods still have an inaccurate pitch and low speaker adaptation quality, there is a significant disparity in pitch between the source and target speaker style domains. As a result, the models tend to generate speech with hoarseness, posing challenges in achieving high-quality voice conversion. In this study, we propose CycleFlow, a novel VC approach that leverages cycle consistency in conditional flow matching (CFM) for speaker timbre adaptation training on non-parallel data. Furthermore, we design a Dual-CFM based on VoiceCFM and PitchCFM to generate speech and improve speaker pitch adaptation quality. Experiments show that our method can significantly improve speaker similarity, generating natural and higher-quality speech.

cs.SD↗

arXiv2024

ESARM: 3D Emotional Speech-to-Animation via Reward Model from Automatically-Ranked Demonstrations

This paper proposes a novel 3D speech-to-animation (STA) generation framework designed to address the shortcomings of existing models in producing diverse and emotionally resonant animations. Current STA models often generate animations that lack emotional depth and variety, failing to align with human expectations. To overcome these limitations, we introduce a novel STA model coupled with a reward model. This combination enables the decoupling of emotion and content under audio conditions through a cross-coupling training approach. Additionally, we develop a training methodology that leverages automatic quality evaluation of generated facial animations to guide the reinforcement learning process. This methodology encourages the STA model to explore a broader range of possibilities, resulting in the generation of diverse and emotionally expressive facial animations of superior quality. We conduct extensive empirical experiments on a benchmark dataset, and the results validate the effectiveness of our proposed framework in generating high-quality, emotionally rich 3D animations that are better aligned with human preferences.

cs.CV↗

arXiv2024

MAIN-VC: Lightweight Speech Representation Disentanglement for One-shot Voice Conversion

One-shot voice conversion aims to change the timbre of any source speech to match that of the unseen target speaker with only one speech sample. Existing methods face difficulties in satisfactory speech representation disentanglement and suffer from sizable networks as some of them leverage numerous complex modules for disentanglement. In this paper, we propose a model named MAIN-VC to effectively disentangle via a concise neural network. The proposed model utilizes Siamese encoders to learn clean representations, further enhanced by the designed mutual information estimator. The Siamese structure and the newly designed convolution module contribute to the lightweight of our model while ensuring performance in diverse voice conversion tasks. The experimental results show that the proposed model achieves comparable subjective scores and exhibits improvements in objective metrics compared to existing methods in a one-shot voice conversion scenario.

cs.SD↗

arXiv2024

IDEAW: Robust Neural Audio Watermarking with Invertible Dual-Embedding

The audio watermarking technique embeds messages into audio and accurately extracts messages from the watermarked audio. Traditional methods develop algorithms based on expert experience to embed watermarks into the time-domain or transform-domain of signals. With the development of deep neural networks, deep learning-based neural audio watermarking has emerged. Compared to traditional algorithms, neural audio watermarking achieves better robustness by considering various attacks during training. However, current neural watermarking methods suffer from low capacity and unsatisfactory imperceptibility. Additionally, the issue of watermark locating, which is extremely important and even more pronounced in neural audio watermarking, has not been adequately studied. In this paper, we design a dual-embedding watermarking model for efficient locating. We also consider the impact of the attack layer on the invertible neural network in robustness training, improving the model to enhance both its reasonableness and stability. Experiments show that the proposed model, IDEAW, can withstand various attacks with higher capacity and more efficient locating ability compared to existing methods.

cs.MM↗

arXiv2024

RREH: Reconstruction Relations Embedded Hashing for Semi-Paired Cross-Modal Retrieval

Known for efficient computation and easy storage, hashing has been extensively explored in cross-modal retrieval. The majority of current hashing models are predicated on the premise of a direct one-to-one mapping between data points. However, in real practice, data correspondence across modalities may be partially provided. In this research, we introduce an innovative unsupervised hashing technique designed for semi-paired cross-modal retrieval tasks, named Reconstruction Relations Embedded Hashing (RREH). RREH assumes that multi-modal data share a common subspace. For paired data, RREH explores the latent consistent information of heterogeneous modalities by seeking a shared representation. For unpaired data, to effectively capture the latent discriminative features, the high-order relationships between unpaired data and anchors are embedded into the latent subspace, which are computed by efficient linear reconstruction. The anchors are sampled from paired data, which improves the efficiency of hash learning. The RREH trains the underlying features and the binary encodings in a unified framework with high-order reconstruction relations preserved. With the well devised objective function and discrete optimization algorithm, RREH is designed to be scalable, making it suitable for large-scale datasets and facilitating efficient cross-modal retrieval. In the evaluation process, the proposed is tested with partially paired data to establish its superiority over several existing methods.

cs.IR↗

arXiv2024

Enhancing Emotion Recognition in Conversation through Emotional Cross-Modal Fusion and Inter-class Contrastive Learning

The purpose of emotion recognition in conversation (ERC) is to identify the emotion category of an utterance based on contextual information. Previous ERC methods relied on simple connections for cross-modal fusion and ignored the information differences between modalities, resulting in the model being unable to focus on modality-specific emotional information. At the same time, the shared information between modalities was not processed to generate emotions. Information redundancy problem. To overcome these limitations, we propose a cross-modal fusion emotion prediction network based on vector connections. The network mainly includes two stages: the multi-modal feature fusion stage based on connection vectors and the emotion classification stage based on fused features. Furthermore, we design a supervised inter-class contrastive learning module based on emotion labels. Experimental results confirm the effectiveness of the proposed method, demonstrating excellent performance on the IEMOCAP and MELD datasets.

cs.CL↗

arXiv2024

RSET: Remapping-based Sorting Method for Emotion Transfer Speech Synthesis

Although current Text-To-Speech (TTS) models are able to generate high-quality speech samples, there are still challenges in developing emotion intensity controllable TTS. Most existing TTS models achieve emotion intensity control by extracting intensity information from reference speeches. Unfortunately, limited by the lack of modeling for intra-class emotion intensity and the model's information decoupling capability, the generated speech cannot achieve fine-grained emotion intensity control and suffers from information leakage issues. In this paper, we propose an emotion transfer TTS model, which defines a remapping-based sorting method to model intra-class relative intensity information, combined with Mutual Information (MI) to decouple speaker and emotion information, and synthesizes expressive speeches with perceptible intensity differences. Experiments show that our model achieves fine-grained emotion control while preserving speaker information.

cs.SD↗

arXiv2024

Learning Expressive Disentangled Speech Representations with Soft Speech Units and Adversarial Style Augmentation

Voice conversion is the task to transform voice characteristics of source speech while preserving content information. Nowadays, self-supervised representation learning models are increasingly utilized in content extraction. However, in these representations, a lot of hidden speaker information leads to timbre leakage while the prosodic information of hidden units lacks use. To address these issues, we propose a novel framework for expressive voice conversion called "SAVC" based on soft speech units from HuBert-soft. Taking soft speech units as input, we design an attribute encoder to extract content and prosody features respectively. Specifically, we first introduce statistic perturbation imposed by adversarial style augmentation to eliminate speaker information. Then the prosody is implicitly modeled on soft speech units with knowledge distillation. Experiment results show that the intelligibility and naturalness of converted speech outperform previous work.

cs.SD↗

arXiv2024

Medical Speech Symptoms Classification via Disentangled Representation

Intent is defined for understanding spoken language in existing works. Both textual features and acoustic features involved in medical speech contain intent, which is important for symptomatic diagnosis. In this paper, we propose a medical speech classification model named DRSC that automatically learns to disentangle intent and content representations from textual-acoustic data for classification. The intent representations of the text domain and the Mel-spectrogram domain are extracted via intent encoders, and then the reconstructed text feature and the Mel-spectrogram feature are obtained through two exchanges. After combining the intent from two domains into a joint representation, the integrated intent representation is fed into a decision layer for classification. Experimental results show that our model obtains an average accuracy rate of 95% in detecting 25 different medical symptoms.

cs.AI↗