Jin Wang

Shichao Dong, Jin Wang, Renhe Ji et al.

In this paper, we analyse the generalization ability of binary classifiers for the task of deepfake detection. We find that the stumbling block to their generalization is caused by the unexpected learned identity representation on images. Termed as the Implicit Identity Leakage, this phenomenon has been qualitatively and quantitatively verified among various DNNs. Furthermore, based on such understanding, we propose a simple yet effective method named the ID-unaware Deepfake Detection Model to reduce the influence of this phenomenon. Extensive experimental results demonstrate that our method outperforms the state-of-the-art in both in-dataset and cross-dataset evaluation. The code is available at https://github.com/megvii-research/CADDM.

Fanqing Meng, Jin Wang, Chuanhao Li et al.

The capability to process multiple images is crucial for Large Vision-Language Models (LVLMs) to develop a more thorough and nuanced understanding of a scene. Recent multi-image LVLMs have begun to address this need. However, their evaluation has not kept pace with their development. To fill this gap, we introduce the Multimodal Multi-image Understanding (MMIU) benchmark, a comprehensive evaluation suite designed to assess LVLMs across a wide range of multi-image tasks. MMIU encompasses 7 types of multi-image relationships, 52 tasks, 77K images, and 11K meticulously curated multiple-choice questions, making it the most extensive benchmark of its kind. Our evaluation of 24 popular LVLMs, including both open-source and proprietary models, reveals significant challenges in multi-image comprehension, particularly in tasks involving spatial understanding. Even the most advanced models, such as GPT-4o, achieve only 55.7% accuracy on MMIU. Through multi-faceted analytical experiments, we identify key performance gaps and limitations, providing valuable insights for future model and data improvements. We aim for MMIU to advance the frontier of LVLM research and development, moving us toward achieving sophisticated multimodal multi-image user interactions.

Ruijun Chen, Jin Wang, Liang-Chih Yu et al.

Maintaining engagement and consistency is particularly important in dialogue systems. Existing works have improved the performance of dialogue systems by intentionally learning interlocutor personas with sophisticated network structures. One issue with this approach is that it requires more personal corpora with annotations. Additionally, these models typically perform the next utterance prediction to generate a response but neglect the discourse coherence in the entire conversation. To address these issues, this study proposes a method of learning to memorize entailment and discourse relations for persona-consistent dialogue tasks. Entailment text pairs in natural language inference dataset were applied to learn latent entailment relations as external memories by premise-to-hypothesis generation task. Furthermore, an internal memory with a similar architecture was applied to the discourse information in the dialogue. Placing orthogonality restrictions on these two memory spaces ensures that the latent entailment relations remain dialogue-independent. Both memories collaborate to obtain entailment and discourse representation for the generation, allowing a deeper understanding of both consistency and coherence. Experiments on two large public datasets, PersonaChat and DSTC7-AVSD, demonstrated the effectiveness of the proposed method. Both automatic and human evaluations indicate that the proposed model outperforms several strong baselines in terms of both persona consistency and response coherence. Our source code is available at https://github.com/Chenrj233/LMEDR.

Bin Liu, Jun Wu, Haoyu Peng et al.

The complex imbalanced label distribution poses a crucial challenge to multi-label classification, as most classifiers are biased towards the majority class and high-frequent labels. Oversampling is an efficient and flexible solution that augments instances to provide a more balanced training dataset for multi-label classifiers. Most existing oversampling methods create synthetic instances in a heuristic way that essentially relies on neighborhood information retrieved using Euclidean distance within the entire feature space. However, they fail to consider the varying semantic relevance of features to different labels, leading to label inconsistency among proximate neighbors and further introducing label confusion and overfitting to synthetic instances. To overcome the above issue, we propose a novel sampling approach called Label-Specific Distance-based Multi-Label Oversampling (LSDMLO) that creates more useful and well-labeled synthetic instances to address the imbalance in multi-label datasets. LSDMLO derives the label-specific distance to identify label-consistent neighbors based on the weighted pertinent feature space, which facilitates selecting seed instances that express more label correlations in boundary areas and generating synthetic instances aligned with the label distribution of original data. The comprehensive experiments verify that the proposed LSDMLO outperforms the state-of-the-art multi-label sampling approaches under various base classifiers.

Mengze Li, Tianbao Wang, Haoyu Zhang et al.

Natural language spatial video grounding aims to detect the relevant objects in video frames with descriptive sentences as the query. In spite of the great advances, most existing methods rely on dense video frame annotations, which require a tremendous amount of human effort. To achieve effective grounding under a limited annotation budget, we investigate one-shot video grounding, and learn to ground natural language in all video frames with solely one frame labeled, in an end-to-end manner. One major challenge of end-to-end one-shot video grounding is the existence of videos frames that are either irrelevant to the language query or the labeled frames. Another challenge relates to the limited supervision, which might result in ineffective representation learning. To address these challenges, we designed an end-to-end model via Information Tree for One-Shot video grounding (IT-OS). Its key module, the information tree, can eliminate the interference of irrelevant frames based on branch search and branch cropping techniques. In addition, several self-supervised tasks are proposed based on the information tree to improve the representation learning under insufficient labeling. Experiments on the benchmark dataset demonstrate the effectiveness of our model.

Xiang Luo, Zhiwen Tang, Jin Wang et al.

Zero-shot dialogue state tracking (DST) seeks to enable dialogue systems to transition to unfamiliar domains without manual annotation or extensive retraining. Prior research has approached this objective by embedding prompts into language models (LMs). Common methodologies include integrating prompts at the input layer or introducing learnable variables at each transformer layer. Nonetheless, each strategy exhibits inherent limitations. Prompts integrated at the input layer risk underutilization, with their impact potentially diminishing across successive transformer layers. Conversely, the addition of learnable variables to each layer can complicate the training process and increase inference latency. To tackle the issues mentioned above, this paper proposes Dual Low-Rank Adaptation (DualLoRA), a plug-and-play architecture designed for zero-shot DST. DualLoRA incorporates two distinct Low-Rank Adaptation (LoRA) components, targeting both dialogue context processing and prompt optimization, to ensure the comprehensive influence of prompts throughout the transformer model layers. This is achieved without incurring additional inference latency, showcasing an efficient integration into existing architectures. Through rigorous evaluation on the MultiWOZ and SGD datasets, DualLoRA demonstrates notable improvements across multiple domains, outperforming traditional baseline methods in zero-shot settings. Our code is accessible at: \url{https://github.com/suntea233/DualLoRA}.

MiroMind Team, Song Bai, Lidong Bing et al.

We present MiroThinker v1.0, an open-source research agent designed to advance tool-augmented reasoning and information-seeking capabilities. Unlike previous agents that only scale up model size or context length, MiroThinker explores interaction scaling at the model level, systematically training the model to handle deeper and more frequent agent-environment interactions as a third dimension of performance improvement. Unlike LLM test-time scaling, which operates in isolation and risks degradation with longer reasoning chains, interactive scaling leverages environment feedback and external information acquisition to correct errors and refine trajectories. Through reinforcement learning, the model achieves efficient interaction scaling: with a 256K context window, it can perform up to 600 tool calls per task, enabling sustained multi-turn reasoning and complex real-world research workflows. Across four representative benchmarks-GAIA, HLE, BrowseComp, and BrowseComp-ZH-the 72B variant achieves up to 81.9%, 37.7%, 47.1%, and 55.6% accuracy respectively, surpassing previous open-source agents and approaching commercial counterparts such as GPT-5-high. Our analysis reveals that MiroThinker benefits from interactive scaling consistently: research performance improves predictably as the model engages in deeper and more frequent agent-environment interactions, demonstrating that interaction depth exhibits scaling behaviors analogous to model size and context length. These findings establish interaction scaling as a third critical dimension for building next-generation open research agents, complementing model capacity and context windows.

Jin Wang, Yu Hu, Lirong Xiang et al.

Computer vision (CV), a non-intrusive and cost-effective technology, has furthered the development of precision livestock farming by enabling optimized decision-making through timely and individualized animal care. The availability of affordable two- and three-dimensional camera sensors, combined with various machine learning and deep learning algorithms, has provided a valuable opportunity to improve livestock production systems. However, despite the availability of various CV tools in the public domain, applying these tools to animal data can be challenging, often requiring users to have programming and data analysis skills, as well as access to computing resources. Moreover, the rapid expansion of precision livestock farming is creating a growing need to educate and train animal science students in CV. This presents educators with the challenge of efficiently demonstrating the complex algorithms involved in CV. Thus, the objective of this study was to develop ShinyAnimalCV, an open-source cloud-based web application. This application provides a user-friendly interface for performing CV tasks, including object segmentation, detection, three-dimensional surface visualization, and extraction of two- and three-dimensional morphological features. Nine pre-trained CV models using top-view animal data are included in the application. ShinyAnimalCV has been deployed online using cloud computing platforms. The source code of ShinyAnimalCV is available on GitHub, along with detailed documentation on training CV models using custom data and deploying ShinyAnimalCV locally to allow users to fully leverage the capabilities of the application. ShinyAnimalCV can contribute to CV research and teaching in the animal science community.

Patrick Bajari, Zhihao Cen, Victor Chernozhukov et al.

We develop empirical models that efficiently process large amounts of unstructured product data (text, images, prices, quantities) to produce accurate hedonic price estimates and derived indices. To achieve this, we generate abstract product attributes (or ``features'') from descriptions and images using deep neural networks. These attributes are then used to estimate the hedonic price function. To demonstrate the effectiveness of this approach, we apply the models to Amazon's data for first-party apparel sales, and estimate hedonic prices. The resulting models have a very high out-of-sample predictive accuracy, with $R^2$ ranging from $80\%$ to $90\%$. Finally, we construct the AI-based hedonic Fisher price index, chained at the year-over-year frequency, and contrast it with the CPI and other electronic indices.

Yueyi Zhang, Jin Wang, Wenming Weng et al.

With the rapid development of deep learning, video deraining has experienced significant progress. However, existing video deraining pipelines cannot achieve satisfying performance for scenes with rain layers of complex spatio-temporal distribution. In this paper, we approach video deraining by employing an event camera. As a neuromorphic sensor, the event camera suits scenes of non-uniform motion and dynamic light conditions. We propose an end-to-end learning-based network to unlock the potential of the event camera for video deraining. First, we devise an event-aware motion detection module to adaptively aggregate multi-frame motion contexts using event-aware masks. Second, we design a pyramidal adaptive selection module for reliably separating the background and rain layers by incorporating multi-modal contextualized priors. In addition, we build a real-world dataset consisting of rainy videos and temporally synchronized event streams. We compare our method with extensive state-of-the-art methods on synthetic and self-collected real-world datasets, demonstrating the clear superiority of our method. The code and dataset are available at \url{https://github.com/booker-max/EGVD}.

Junzhuo Liu, Qiaosong Chen, Ye Zhang et al.

Clinically, automated polyp segmentation techniques have the potential to significantly improve the efficiency and accuracy of medical diagnosis, thereby reducing the risk of colorectal cancer in patients. Unfortunately, existing methods suffer from two significant weaknesses that can impact the accuracy of segmentation. Firstly, features extracted by encoders are not adequately filtered and utilized. Secondly, semantic conflicts and information redundancy caused by feature fusion are not attended to. To overcome these limitations, we propose a novel approach for polyp segmentation, named MLFF-Net, which leverages multi-level feature fusion and attention mechanisms. Specifically, MLFF-Net comprises three modules: Multi-scale Attention Module (MAM), High-level Feature Enhancement Module (HFEM), and Global Attention Module (GAM). Among these, MAM is used to extract multi-scale information and polyp details from the shallow output of the encoder. In HFEM, the deep features of the encoders complement each other by aggregation. Meanwhile, the attention mechanism redistributes the weight of the aggregated features, weakening the conflicting redundant parts and highlighting the information useful to the task. GAM combines features from the encoder and decoder features, as well as computes global dependencies to prevent receptive field locality. Experimental results on five public datasets show that the proposed method not only can segment multiple types of polyps but also has advantages over current state-of-the-art methods in both accuracy and generalization ability.

Bin Liu, Jin Wang, Kaiwei Sun et al.

The discovery of drug-target interactions (DTIs) is a pivotal process in pharmaceutical development. Computational approaches are a promising and efficient alternative to tedious and costly wet-lab experiments for predicting novel DTIs from numerous candidates. Recently, with the availability of abundant heterogeneous biological information from diverse data sources, computational methods have been able to leverage multiple drug and target similarities to boost the performance of DTI prediction. Similarity integration is an effective and flexible strategy to extract crucial information across complementary similarity views, providing a compressed input for any similarity-based DTI prediction model. However, existing similarity integration methods filter and fuse similarities from a global perspective, neglecting the utility of similarity views for each drug and target. In this study, we propose a Fine-Grained Selective similarity integration approach, called FGS, which employs a local interaction consistency-based weight matrix to capture and exploit the importance of similarities at a finer granularity in both similarity selection and combination steps. We evaluate FGS on five DTI prediction datasets under various prediction settings. Experimental results show that our method not only outperforms similarity integration competitors with comparable computational costs, but also achieves better prediction performance than state-of-the-art DTI prediction approaches by collaborating with conventional base models. Furthermore, case studies on the analysis of similarity weights and on the verification of novel predictions confirm the practical ability of FGS.

Shichao Dong, Jin Wang, Jiajun Liang et al.

This paper aims to interpret how deepfake detection models learn artifact features of images when just supervised by binary labels. To this end, three hypotheses from the perspective of image matching are proposed as follows. 1. Deepfake detection models indicate real/fake images based on visual concepts that are neither source-relevant nor target-relevant, that is, considering such visual concepts as artifact-relevant. 2. Besides the supervision of binary labels, deepfake detection models implicitly learn artifact-relevant visual concepts through the FST-Matching (i.e. the matching fake, source, target images) in the training set. 3. Implicitly learned artifact visual concepts through the FST-Matching in the raw training set are vulnerable to video compression. In experiments, the above hypotheses are verified among various DNNs. Furthermore, based on this understanding, we propose the FST-Matching Deepfake Detection Model to boost the performance of forgery detection on compressed videos. Experiment results show that our method achieves great performance, especially on highly-compressed (e.g. c40) videos.

Li Yuan, Yi Cai, Jin Wang et al.

Multimodal named entity recognition (MNER) and multimodal relation extraction (MRE) are two fundamental subtasks in the multimodal knowledge graph construction task. However, the existing methods usually handle two tasks independently, which ignores the bidirectional interaction between them. This paper is the first to propose jointly performing MNER and MRE as a joint multimodal entity-relation extraction task (JMERE). Besides, the current MNER and MRE models only consider aligning the visual objects with textual entities in visual and textual graphs but ignore the entity-entity relationships and object-object relationships. To address the above challenges, we propose an edge-enhanced graph alignment network and a word-pair relation tagging (EEGA) for JMERE task. Specifically, we first design a word-pair relation tagging to exploit the bidirectional interaction between MNER and MRE and avoid the error propagation. Then, we propose an edge-enhanced graph alignment network to enhance the JMERE task by aligning nodes and edges in the cross-graph. Compared with previous methods, the proposed method can leverage the edge information to auxiliary alignment between objects and entities and find the correlations between entity-entity relationships and object-object relationships. Experiments are conducted to show the effectiveness of our model.

Sheng Zhang, Jin Wang, Haitao Jiang et al.

With the growing popularity of deep-learning models, model understanding becomes more important. Much effort has been devoted to demystify deep neural networks for better interpretability. Some feature attribution methods have shown promising results in computer vision, especially the gradient-based methods where effectively smoothing the gradients with reference data is key to a robust and faithful result. However, direct application of these gradient-based methods to NLP tasks is not trivial due to the fact that the input consists of discrete tokens and the "reference" tokens are not explicitly defined. In this work, we propose Locally Aggregated Feature Attribution (LAFA), a novel gradient-based feature attribution method for NLP models. Instead of relying on obscure reference tokens, it smooths gradients by aggregating similar reference texts derived from language model embeddings. For evaluation purpose, we also design experiments on different NLP tasks including Entity Recognition and Sentiment Analysis on public datasets as well as key feature detection on a constructed Amazon catalogue dataset. The superior performance of the proposed method is demonstrated through experiments.

Jin Wang, Arturo Laurenzi, Nikos Tsagarakis

Enabling humanoid robots to perform autonomously loco-manipulation in unstructured environments is crucial and highly challenging for achieving embodied intelligence. This involves robots being able to plan their actions and behaviors in long-horizon tasks while using multi-modality to perceive deviations between task execution and high-level planning. Recently, large language models (LLMs) have demonstrated powerful planning and reasoning capabilities for comprehension and processing of semantic information through robot control tasks, as well as the usability of analytical judgment and decision-making for multi-modal inputs. To leverage the power of LLMs towards humanoid loco-manipulation, we propose a novel language-model based framework that enables robots to autonomously plan behaviors and low-level execution under given textual instructions, while observing and correcting failures that may occur during task execution. To systematically evaluate this framework in grounding LLMs, we created the robot 'action' and 'sensing' behavior library for task planning, and conducted mobile manipulation tasks and experiments in both simulated and real environments using the CENTAURO robot, and verified the effectiveness and application of this approach in robotic tasks with autonomous behavioral planning.

Zerun Liu, Fan Zhang, Jingxuan He et al.

In the realm of multi-modality, text-guided image retouching techniques emerged with the advent of deep learning. Most currently available text-guided methods, however, rely on object-level supervision to constrain the region that may be modified. This not only makes it more challenging to develop these algorithms, but it also limits how widely deep learning can be used for image retouching. In this paper, we offer a text-guided mask-free image retouching approach that yields consistent results to address this concern. In order to perform image retouching without mask supervision, our technique can construct plausible and edge-sharp masks based on the text for each object in the image. Extensive experiments have shown that our method can produce high-quality, accurate images based on spoken language. The source code will be released soon.

Cheng Ge, Xi Chen, Ming Wang et al.

Deep learning has seen increasing applications in time series in recent years. For time series anomaly detection scenarios, such as in finance, Internet of Things, data center operations, etc., time series usually show very flexible baselines depending on various external factors. Anomalies unveil themselves by lying far away from the baseline. However, the detection is not always easy due to some challenges including baseline shifting, lacking of labels, noise interference, real time detection in streaming data, result interpretability, etc. In this paper, we develop a novel deep architecture to properly extract the baseline from time series, namely Deep Baseline Network (DBLN). By using this deep network, we can easily locate the baseline position and then provide reliable and interpretable anomaly detection result. Empirical evaluation on both synthetic and public real-world datasets shows that our purely unsupervised algorithm achieves superior performance compared with state-of-art methods and has good practical applications.

Kai Cheng, Xiaoxiao Long, Wei Yin et al.

Multi-camera setups find widespread use across various applications, such as autonomous driving, as they greatly expand sensing capabilities. Despite the fast development of Neural radiance field (NeRF) techniques and their wide applications in both indoor and outdoor scenes, applying NeRF to multi-camera systems remains very challenging. This is primarily due to the inherent under-calibration issues in multi-camera setup, including inconsistent imaging effects stemming from separately calibrated image signal processing units in diverse cameras, and system errors arising from mechanical vibrations during driving that affect relative camera poses. In this paper, we present UC-NeRF, a novel method tailored for novel view synthesis in under-calibrated multi-view camera systems. Firstly, we propose a layer-based color correction to rectify the color inconsistency in different image regions. Second, we propose virtual warping to generate more viewpoint-diverse but color-consistent virtual views for color correction and 3D recovery. Finally, a spatiotemporally constrained pose refinement is designed for more robust and accurate pose calibration in multi-camera systems. Our method not only achieves state-of-the-art performance of novel view synthesis in multi-camera setups, but also effectively facilitates depth estimation in large-scale outdoor scenes with the synthesized novel views.

Ye Bi, Leticia M. Campos, Jin Wang et al.

Monitoring cow body weight is crucial to support farm management decisions due to its direct relationship with the growth, nutritional status, and health of dairy cows. Cow body weight is a repeated trait, however, the majority of previous body weight prediction research only used data collected at a single point in time. Furthermore, the utility of deep learning-based segmentation for body weight prediction using videos remains unanswered. Therefore, the objectives of this study were to predict cow body weight from repeatedly measured video data, to compare the performance of the thresholding and Mask R-CNN deep learning approaches, to evaluate the predictive ability of body weight regression models, and to promote open science in the animal science community by releasing the source code for video-based body weight prediction. A total of 40,405 depth images and depth map files were obtained from 10 lactating Holstein cows and 2 non-lactating Jersey cows. Three approaches were investigated to segment the cow's body from the background, including single thresholding, adaptive thresholding, and Mask R-CNN. Four image-derived biometric features, such as dorsal length, abdominal width, height, and volume, were estimated from the segmented images. On average, the Mask-RCNN approach combined with a linear mixed model resulted in the best prediction coefficient of determination and mean absolute percentage error of 0.98 and 2.03%, respectively, in the forecasting cross-validation. The Mask-RCNN approach was also the best in the leave-three-cows-out cross-validation. The prediction coefficients of determination and mean absolute percentage error of the Mask-RCNN coupled with the linear mixed model were 0.90 and 4.70%, respectively. Our results suggest that deep learning-based segmentation improves the prediction performance of cow body weight from longitudinal depth video data.

Jin Wang, Liang Lin, Kaiwen Luo et al.

While Audio Large Language Models (ALLMs) have achieved remarkable progress in understanding and generation, their potential privacy implications remain largely unexplored. This paper takes the first step to investigate whether ALLMs inadvertently leak user privacy solely through acoustic voiceprints and introduces $\textit{HearSay}$, a comprehensive benchmark constructed from over 22,000 real-world audio clips. To ensure data quality, the benchmark is meticulously curated through a rigorous pipeline involving automated profiling and human verification, guaranteeing that all privacy labels are grounded in factual records. Extensive experiments on $\textit{HearSay}$ yield three critical findings: $\textbf{Significant Privacy Leakage}$: ALLMs inherently extract private attributes from voiceprints, reaching 92.89% accuracy on gender and effectively profiling social attributes. $\textbf{Insufficient Safety Mechanisms}$: Alarmingly, existing safeguards are severely inadequate; most models fail to refuse privacy-intruding requests, exhibiting near-zero refusal rates for physiological traits. $\textbf{Reasoning Amplifies Risk}$: Chain-of-Thought (CoT) reasoning exacerbates privacy risks in capable models by uncovering deeper acoustic correlations. These findings expose critical vulnerabilities in ALLMs, underscoring the urgent need for targeted privacy alignment. The codes and dataset are available at https://github.com/JinWang79/HearSay_Benchmark

Kaining Ying, Fanqing Meng, Jin Wang et al.

Large Vision-Language Models (LVLMs) show significant strides in general-purpose multimodal applications such as visual dialogue and embodied navigation. However, existing multimodal evaluation benchmarks cover a limited number of multimodal tasks testing rudimentary capabilities, falling short in tracking LVLM development. In this study, we present MMT-Bench, a comprehensive benchmark designed to assess LVLMs across massive multimodal tasks requiring expert knowledge and deliberate visual recognition, localization, reasoning, and planning. MMT-Bench comprises $31,325$ meticulously curated multi-choice visual questions from various multimodal scenarios such as vehicle driving and embodied navigation, covering $32$ core meta-tasks and $162$ subtasks in multimodal understanding. Due to its extensive task coverage, MMT-Bench enables the evaluation of LVLMs using a task map, facilitating the discovery of in- and out-of-domain tasks. Evaluation results involving $30$ LVLMs such as the proprietary GPT-4V, GeminiProVision, and open-sourced InternVL-Chat, underscore the significant challenges posed by MMT-Bench. We anticipate that MMT-Bench will inspire the community to develop next-generation multimodal foundation models aimed at achieving general-purpose multimodal intelligence.

Jie Wang, Jin Wang, Xuejie Zhang

Metaphors are common in everyday language, and the identification and understanding of metaphors are facilitated by models to achieve a better understanding of the text. Metaphors are mainly identified and generated by pre-trained models in existing research, but situations, where tenors or vehicles are not included in the metaphor, cannot be handled. The problem can be effectively solved by using Large Language Models (LLMs), but significant room for exploration remains in this early-stage research area. A multi-stage generative heuristic-enhanced prompt framework is proposed in this study to enhance the ability of LLMs to recognize tenors, vehicles, and grounds in Chinese metaphors. In the first stage, a small model is trained to obtain the required confidence score for answer candidate generation. In the second stage, questions are clustered and sampled according to specific rules. Finally, the heuristic-enhanced prompt needed is formed by combining the generated answer candidates and demonstrations. The proposed model achieved 3rd place in Track 1 of Subtask 1, 1st place in Track 2 of Subtask 1, and 1st place in both tracks of Subtask 2 at the NLPCC-2024 Shared Task 9.

Yitian Gong, Botian Jiang, Yiwei Zhao et al.

This technical report presents MOSS-TTS, a speech generation foundation model built on a scalable recipe: discrete audio tokens, autoregressive modeling, and large-scale pretraining. Built on MOSS-Audio-Tokenizer, a causal Transformer tokenizer that compresses 24 kHz audio to 12.5 fps with variable-bitrate RVQ and unified semantic-acoustic representations, we release two complementary generators: MOSS-TTS, which emphasizes structural simplicity, scalability, and long-context/control-oriented deployment, and MOSS-TTS-Local-Transformer, which introduces a frame-local autoregressive module for higher modeling efficiency, stronger speaker preservation, and a shorter time to first audio. Across multilingual and open-domain settings, MOSS-TTS supports zero-shot voice cloning, token-level duration control, phoneme-/pinyin-level pronunciation control, smooth code-switching, and stable long-form generation. This report summarizes the design, training recipe, and empirical characteristics of the released models.

Jin Wang, Wenbin Jiang, Xiangbo Wang et al.

Neural audio compression has emerged as a promising technology for efficiently representing speech, music, and general audio. However, existing methods suffer from significant performance degradation at limited bitrates, where the available embedding space is sharply constrained. To address this, we propose a universal high-fidelity neural audio compression algorithm featuring Residual Experts Vector Quantization (REVQ), which substantially expands the embedding space with minimal impact on bandwidth. A gentle load-balancing strategy is introduced to ensure the full utilization of this expanded space. Furthermore, we develop a novel multi-tiered discriminator that periodically stratifies STFT spectra, guiding the generator to focus on critical spectral regions. To support multiple bitrates without quality loss at the lower end, we adopt an efficient post-training strategy. Our proposed model achieves impressive performance, with PESQ and ViSQOL scores of 2.87 and 4.27, respectively, at 2.67 kbps bandwidth. The approach effectively reduces spectral blur, decreasing the distance to the original mel-spectrogram by 13%. Notably, our post-training strategy achieves performance comparable to dedicated fixed-bitrate models while reducing the required training time by half. Extensive ablation studies confirm the superiority of our method over baselines.

Kewei Zhang, Jin Wang, Sensen Gao et al.

End-to-end autonomous driving via Vision-Language-Action (VLA) models demands a precarious balance between high-fidelity trajectory planning and efficient inference. Existing paradigms typically fall short: autoregressive (AR) VLAs are memory-bandwidth-bound on edge hardware and prone to exposure-bias drift, while full-sequence diffusion models preclude KV-cache reuse and suffer from "logical leakage" that violates the fundamental perceive-then-plan causality. We present Fast-dDrive, a block-diffusion VLA that performs bidirectional refinement within semantic units while enforcing strict causal ordering across them. Leveraging the observation that driving VLAs often emit structured JSON-like outputs, Fast-dDrive freezes structural tokens into a section scaffold and employs a section-aware training recipe that prioritizes safety-critical planning. We further introduce Scaffold Speculative Decoding to achieve AR-equivalent quality at significantly higher throughput. Finally, we propose a low-overhead test-time scaling scheme: by forking $N$ stochastic trajectory rollouts from a single shared-prefix KV cache and averaging them, we effectively suppress prediction variance at a fractional computational cost. Empirical results demonstrate that Fast-dDrive redefines the speed-accuracy frontier for driving agents. On the WOD-E2E test set, Fast-dDrive achieves SOTA ADE@3s and ADE@5s, alongside the highest RFS among diffusion-based VLAs; on nuScenes, it reduces average L2 error to $0.32$m (a $22\%$ improvement). When integrated with SGLang, our framework delivers $12\times$ throughput speedup over the AR baseline, narrowing the gap between high-capacity VLAs and the efficiency demands of real-time on-vehicle deployment.

Chengyue Wu, Shiyi Lan, Yonggan Fu et al.

Vision-language models (VLMs) predominantly rely on autoregressive decoding, which generates tokens one at a time and fundamentally limits inference throughput. This limitation is especially acute in physical AI scenarios such as robotics and autonomous driving, where VLMs are deployed on edge devices at batch size one, making AR decoding memory-bandwidth-bound and leaving hardware parallelism underutilized. While block-wise discrete diffusion has shown promise for parallel text generation, extending it to VLMs remains challenging due to the need to jointly handle continuous visual representations and discrete text tokens while preserving pretrained multimodal capabilities. We present Fast-dVLM, a block-diffusion-based VLM that enables KV-cache-compatible parallel decoding and speculative block decoding for inference acceleration. We systematically compare two AR-to-diffusion conversion strategies: a two-stage approach that first adapts the LLM backbone with text-only diffusion fine-tuning before multimodal training, and a direct approach that converts the full AR VLM in one stage. Under comparable training budgets, direct conversion proves substantially more efficient by leveraging the already multimodally aligned VLM; we therefore adopt it as our recommended recipe. We introduce a suite of multimodal diffusion adaptations, block size annealing, causal context attention, auto-truncation masking, and vision efficient concatenation, that collectively enable effective block diffusion in the VLM setting. Extensive experiments across 11 multimodal benchmarks show Fast-dVLM matches its autoregressive counterpart in generation quality. With SGLang integration and FP8 quantization, Fast-dVLM achieves over 6x end-to-end inference speedup over the AR baseline.

Yucheng Shi, Ying Li, Yu Wang et al.

Large language models (LLMs) are promising backbones for generative recommender systems, yet a key challenge remains underexplored: verbalization, i.e., converting structured user interaction logs into effective natural language inputs. Existing methods rely on rigid templates that simply concatenate fields, yielding suboptimal representations for recommendation. We propose a data-centric framework that learns verbalization for LLM-based recommendation. Using reinforcement learning, a verbalization agent transforms raw interaction histories into optimized textual contexts, with recommendation accuracy as the training signal. This agent learns to filter noise, incorporate relevant metadata, and reorganize information to improve downstream predictions. Experiments on a large-scale industrial streaming dataset show that learned verbalization delivers up to 93% relative improvement in discovery item recommendation accuracy over template-based baselines. Further analysis reveals emergent strategies such as user interest summarization, noise removal, and syntax normalization, offering insights into effective context construction for LLM-based recommender systems.

Jonas Becker, Liang-Chih Yu, Shamsuddeen Hassan Muhammad et al.

Stance detection is an established task that classifies an author's attitude toward a specific target into categories such as Favor, Neutral, and Against. Beyond categorical stance labels, we leverage a long-established affective science framework to model stance along real-valued dimensions of valence (negative-positive) and arousal (calm-active). This dimensional approach captures nuanced affective states underlying stance expressions, enabling fine-grained stance analysis. To this end, we introduce DimStance, the first dimensional stance resource with valence-arousal (VA) annotations. This resource comprises 11,746 target aspects in 7,365 texts across five languages (English, German, Chinese, Nigerian Pidgin, and Swahili) and two domains (politics and environmental protection). To facilitate the evaluation of stance VA prediction, we formulate the dimensional stance regression task, analyze cross-lingual VA patterns, and benchmark pretrained and large language models under regression and prompting settings. Results show competitive performance of fine-tuned LLM regressors, persistent challenges in low-resource languages, and limitations of token-based generation. DimStance provides a foundation for multilingual, emotion-aware, stance analysis and benchmarking.

Shurong Yang, Huadong Li, Juhao Wu et al.

Diffusion models have demonstrated superior performance in the field of portrait animation. However, current approaches relied on either visual or audio modality to control character movements, failing to exploit the potential of mixed-modal control. This challenge arises from the difficulty in balancing the weak control strength of audio modality and the strong control strength of visual modality. To address this issue, we introduce MegActor-$Σ$: a mixed-modal conditional diffusion transformer (DiT), which can flexibly inject audio and visual modality control signals into portrait animation. Specifically, we make substantial advancements over its predecessor, MegActor, by leveraging the promising model structure of DiT and integrating audio and visual conditions through advanced modules within the DiT framework. To further achieve flexible combinations of mixed-modal control signals, we propose a ``Modality Decoupling Control" training strategy to balance the control strength between visual and audio modalities, along with the ``Amplitude Adjustment" inference strategy to freely regulate the motion amplitude of each modality. Finally, to facilitate extensive studies in this field, we design several dataset evaluation metrics to filter out public datasets and solely use this filtered dataset to train MegActor-$Σ$. Extensive experiments demonstrate the superiority of our approach in generating vivid portrait animations, outperforming previous methods trained on private dataset.

Tiesunlong Shen, Rui Mao, Jin Wang et al.

Aligning Large Language Models (LLMs) with human preferences is critical, yet traditional fine-tuning methods are computationally expensive and inflexible. While test-time alignment offers a promising alternative, existing approaches often rely on distorted trajectory-level signals or inefficient sampling, fundamentally capping performance and failing to preserve the generative diversity of the base model. This paper introduces LLMdoctor, a novel framework for efficient test-time alignment that operates via a patient-doctor paradigm. It integrates token-level reward acquisition with token-level flow-guided preference optimization (TFPO) to steer a large, frozen patient LLM with a smaller, specialized doctor model. Unlike conventional methods that rely on trajectory-level rewards, LLMdoctor first extracts fine-grained, token-level preference signals from the patient model's behavioral variations. These signals then guide the training of the doctor model via TFPO, which establishes flow consistency across all subtrajectories, enabling precise token-by-token alignment while inherently preserving generation diversity. Extensive experiments demonstrate that LLMdoctor significantly outperforms existing test-time alignment methods and even surpasses the performance of full fine-tuning approaches like DPO.

Kaiyuan Chen, Guangmin Zheng, Jin Wang et al.

Existing self-evolution methods overlook the influence of fine-grained reasoning steps, which leads to the reasoner-verifier gap. The computational inefficiency of Monte Carlo (MC) process supervision further exacerbates the difficulty in mitigating the gap. Motivated by the Error-Related Negativity (ERN), which the reasoner can localize error following incorrect decisions, guiding rapid adjustments, we propose a Self-Adaptive Process Optimization (SAPO) method for self-improvement in Small Language Models (SLMs). SAPO adaptively and efficiently introduces process supervision signals by actively minimizing the reasoner-verifier gap rather than relying on inefficient MC estimations. Extensive experiments demonstrate that the proposed method outperforms most existing self-evolution methods on two challenging task types: mathematics and code. Additionally, to further investigate SAPO's impact on verifier performance, this work introduces two new benchmarks for process reward models in both mathematical and coding tasks.

Keqin Li, Jin Wang, Xubo Wu et al.

With the rapid growth of global e-commerce, the demand for automation in the logistics industry is increasing. This study focuses on automated picking systems in warehouses, utilizing deep learning and reinforcement learning technologies to enhance picking efficiency and accuracy while reducing system failure rates. Through empirical analysis, we demonstrate the effectiveness of these technologies in improving robot picking performance and adaptability to complex environments. The results show that the integrated machine learning model significantly outperforms traditional methods, effectively addressing the challenges of peak order processing, reducing operational errors, and improving overall logistics efficiency. Additionally, by analyzing environmental factors, this study further optimizes system design to ensure efficient and stable operation under variable conditions. This research not only provides innovative solutions for logistics automation but also offers a theoretical and empirical foundation for future technological development and application.

Jian Pang, Bingfeng Zhang, Jin Wang et al.

Object detection in hazy environments is challenging because degraded objects are nearly invisible and their semantics are weakened by environmental noise, making it difficult for detectors to identify. Common approaches involve image enhancement to boost weakened semantics, but these methods are limited by the instability of enhanced modules. This paper proposes a novel solution by employing language prompts to enhance weakened semantics without image enhancement. Specifically, we design Approximation of Mutual Exclusion (AME) to provide credible weights for Cross-Entropy Loss, resulting in CLIP-guided Cross-Entropy Loss (CLIP-CE). The provided weights assess the semantic weakening of objects. Through the backpropagation of CLIP-CE, weakened semantics are enhanced, making degraded objects easier to detect. In addition, we present Fine-tuned AME (FAME) which adaptively fine-tunes the weight of AME based on the predicted confidence. The proposed FAME compensates for the imbalanced optimization in AME. Furthermore, we present HazyCOCO, a large-scale synthetic hazy dataset comprising 61258 images. Experimental results demonstrate that our method achieves state-of-the-art performance. The code and dataset will be released.

Jeffrey J. Donatelli, Miaoqi Chu, Zixi Hu et al.

Coherent surface scattering imaging (CSSI) is an emerging experimental technique uniquely suited to probing the structure of thin nanostructures. In these experiments, a specimen is placed on a substrate, and a series of X-ray diffraction patterns is collected at grazing incidence angles as the specimen is rotated. However, reconstructing the specimen's 3D structure from the data is challenging due to dynamical scattering effects induced by the experimental geometry and the lack of direct phase measurements. Specifically, the data involves nonuniformly sampled Fourier-transform values of the specimen density, and failure to effectively address this nonuniformity can lead to errors or degraded performance. Here we introduce a mathematical inversion framework that combines iterative-projection-based phasing techniques with new fast nonuniform Fourier inversion methods to efficiently reconstruct isolated 3D structures from their CSSI rotation-series data. We also analyze the theoretical properties of CSSI reconstruction to derive requirements on experimental parameters and characterize solution uniqueness. We validate our approach using CSSI data simulated from a conical Siemens star and a porous medium, demonstrating that high-resolution 3D structures can be reconstructed even in the presence of significant dynamical scattering, from data collected at as few as one or two incident angles. More broadly, the presented nonuniform reconstruction framework provides a foundation for solving challenging generalizations of the phase problem in which measurements involve nonlinear combinations of nonuniformly sampled Fourier values.

Lung-Hao Lee, Liang-Chih Yu, Natalia Loukashevich et al.

Aspect-Based Sentiment Analysis (ABSA) focuses on extracting sentiment at a fine-grained aspect level and has been widely applied across real-world domains. However, existing ABSA research relies on coarse-grained categorical labels (e.g., positive, negative), which limits its ability to capture nuanced affective states. To address this limitation, we adopt a dimensional approach that represents sentiment with continuous valence-arousal (VA) scores, enabling fine-grained analysis at both the aspect and sentiment levels. To this end, we introduce DimABSA, the first multilingual, dimensional ABSA resource annotated with both traditional ABSA elements (aspect terms, aspect categories, and opinion terms) and newly introduced VA scores. This resource contains 76,958 aspect instances across 42,590 sentences, spanning six languages and four domains. We further introduce three subtasks that combine VA scores with different ABSA elements, providing a bridge from traditional ABSA to dimensional ABSA. Given that these subtasks involve both categorical and continuous outputs, we propose a new unified metric, continuous F1 (cF1), which incorporates VA prediction error into standard F1. We provide a comprehensive benchmark using both prompted and fine-tuned large language models across all subtasks. Our results show that DimABSA is a challenging benchmark and provides a foundation for advancing multilingual dimensional ABSA.

You Zhang, Jin Wang, Liang-Chih Yu et al.

Effectively and efficiently adapting a pre-trained language model (PLM) for human-centered text understanding (HCTU) is challenging since user tokens are million-level in most personalized applications and do not have concrete explicit semantics. A standard and parameter-efficient approach (e.g., LoRA) necessitates memorizing numerous suits of adapters for each user. In this work, we introduce a personalized LoRA (PLoRA) with a plug-and-play (PnP) framework for the HCTU task. PLoRA is effective, parameter-efficient, and dynamically deploying in PLMs. Moreover, a personalized dropout and a mutual information maximizing strategies are adopted and hence the proposed PLoRA can be well adapted to few/zero-shot learning scenarios for the cold-start issue. Experiments conducted on four benchmark datasets show that the proposed method outperforms existing methods in full/few/zero-shot learning scenarios for the HCTU task, even though it has fewer trainable parameters. For reproducibility, the code for this paper is available at: https://github.com/yoyo-yun/PLoRA.

Jin Wang, Changlin Song

Artificial intelligence has made significant progress in the Close World problem, being able to accurately recognize old knowledge through training and classification. However, AI faces significant challenges in the Open World problem, as it involves a new and unknown exploration journey. AI is not inherently proactive in exploration, and its challenge lies in not knowing how to approach and adapt to the unknown world. How do humans acquire knowledge of the unknown world. Humans identify new knowledge through intrinsic cognition. In the process of recognizing new colors, the cognitive cues are different from known color features and involve hue, saturation, brightness, and other characteristics. When AI encounters objects with different features in the new world, it faces another challenge: where are the distinguishing features between influential features of new and old objects? AI often mistakes a new world's brown bear for a known dog because it has not learned the differences in feature distributions between knowledge systems. This is because things in the new and old worlds have different units and dimensions for their features. This paper proposes an open-world model and elemental feature system that focuses on fundamentally recognizing the distribution differences in objective features between the new and old worlds. The quantum tunneling effect of learning ability in the new and old worlds is realized through the tractive force of meta-characteristic. The outstanding performance of the model system in learning new knowledge (using pedestrian re-identification datasets as an example) demonstrates that AI has acquired the ability to recognize the new world with an accuracy of $96.71\%$ at most and has gained the capability to explore new knowledge, similar to humans.

Jin Wang, Liang-Chih Yu, Xuejie Zhang

The pre-training for language models captures general language understanding but fails to distinguish the affective impact of a particular context to a specific word. Recent works have sought to introduce contrastive learning (CL) for sentiment-aware pre-training in acquiring affective information. Nevertheless, these methods present two significant limitations. First, the compatibility of the GPU memory often limits the number of negative samples, hindering the opportunities to learn good representations. In addition, using only a few sentiment polarities as hard labels, e.g., positive, neutral, and negative, to supervise CL will force all representations to converge to a few points, leading to the issue of latent space collapse. This study proposes a soft momentum contrastive learning (SoftMCL) for fine-grained sentiment-aware pre-training. Instead of hard labels, we introduce valence ratings as soft-label supervision for CL to fine-grained measure the sentiment similarities between samples. The proposed SoftMCL is conducted on both the word- and sentence-level to enhance the model's ability to learn affective information. A momentum queue was introduced to expand the contrastive samples, allowing storing and involving more negatives to overcome the limitations of hardware platforms. Extensive experiments were conducted on four different sentiment-related tasks, which demonstrates the effectiveness of the proposed SoftMCL method. The code and data of the proposed SoftMCL is available at: https://www.github.com/wangjin0818/SoftMCL/.

Xiang Luo, Zhiwen Tang, Jin Wang et al.

User Simulators play a pivotal role in training and evaluating task-oriented dialogue systems. Traditional user simulators typically rely on human-engineered agendas, resulting in generated responses that often lack diversity and spontaneity. Although large language models (LLMs) exhibit a remarkable capacity for generating coherent and contextually appropriate utterances, they may fall short when tasked with generating responses that effectively guide users towards their goals, particularly in dialogues with intricate constraints and requirements. This paper introduces DuetSim, a novel framework designed to address the intricate demands of task-oriented dialogues by leveraging LLMs. DuetSim stands apart from conventional approaches by employing two LLMs in tandem: one dedicated to response generation and the other focused on verification. This dual LLM approach empowers DuetSim to produce responses that not only exhibit diversity but also demonstrate accuracy and are preferred by human users. We validate the efficacy of our method through extensive experiments conducted on the MultiWOZ dataset, highlighting improvements in response quality and correctness, largely attributed to the incorporation of the second LLM. Our code is accessible at: https://github.com/suntea233/DuetSim.

Chang Gong, Di Yao, Jin Wang et al.

Root Cause Analysis (RCA) aims at identifying the underlying causes of system faults by uncovering and analyzing the causal structure from complex systems. It has been widely used in many application domains. Reliable diagnostic conclusions are of great importance in mitigating system failures and financial losses. However, previous studies implicitly assume a full observation of the system, which neglect the effect of partial observation (i.e., missing nodes and latent malfunction). As a result, they fail in deriving reliable RCA results. In this paper, we unveil the issues of unobserved confounders and heterogeneity in partial observation and come up with a new problem of root cause analysis with partially observed data. To achieve this, we propose PORCA, a novel RCA framework which can explore reliable root causes under both unobserved confounders and unobserved heterogeneity. PORCA leverages magnified score-based causal discovery to efficiently optimize acyclic directed mixed graph under unobserved confounders. In addition, we also develop a heterogeneity-aware scheduling strategy to provide adaptive sample weights. Extensive experimental results on one synthetic and two real-world datasets demonstrate the effectiveness and superiority of the proposed framework.

Kelu Yao, Nuo Xu, Rong Yang et al.

This paper introduces a holistic vision-language foundation model tailored for remote sensing, named Falcon. Falcon offers a unified, prompt-based paradigm that effectively executes comprehensive and complex remote sensing tasks. Falcon demonstrates powerful understanding and reasoning abilities at the image, region, and pixel levels. Specifically, given simple natural language instructions and remote sensing images, Falcon can produce impressive results in text form across 14 distinct tasks, i.e., image classification, object detection, segmentation, image captioning, and etc. To facilitate Falcon's training and empower its representation capacity to encode rich spatial and semantic information, we developed Falcon_SFT, a large-scale, multi-task, instruction-tuning dataset in the field of remote sensing. The Falcon_SFT dataset consists of approximately 78 million high-quality data samples, covering 5.6 million multi-spatial resolution and multi-view remote sensing images with diverse instructions. It features hierarchical annotations and undergoes manual sampling verification to ensure high data quality and reliability. Extensive comparative experiments are conducted, which verify that Falcon achieves remarkable performance over 67 datasets and 14 tasks, despite having only 0.7B parameters. We release the complete dataset, code, and model weights at https://github.com/TianHuiLab/Falcon, hoping to help further develop the open-source community.

Zesheng Jia, Jin Wang, Siao Liu et al.

Fast domain adaptation remains a fundamental challenge for deploying multi-agent systems across diverse environments in Vehicle-to-Everything (V2X) collaborative perception. Despite the success of Parameter-Efficient Fine-Tuning (PEFT) in natural language processing and conventional vision tasks, directly applying PEFT to multi-agent settings leads to significant performance degradation and training instability. In this work, we conduct a detailed analysis and identify two key factors: (i) inter-frame redundancy in heterogeneous sensory streams, and (ii) erosion of fine-grained semantics in deep-layer representations under PEFT adaptation. To address these issues, we propose FlowAdapt, a parameter-efficient framework grounded in optimal transport theory, which minimizes information transport costs across both data distributions and network hierarchies. Specifically, we introduce a Wasserstein Greedy Sampling strategy to selectively filter redundant samples via a bounded covering radius. Furthermore, Progressive Knowledge Transfer module is designed to progressively inject compressed early-stage representations into later stages through learnable pathways, alleviating semantic degradation in late-stage adaptation. Extensive experiments on three benchmarks demonstrate that FlowAdapt achieves state-of-the-art performance with only 1% of trainable parameters, effectively bridging domain gaps with superior sample efficiency and generalization.

Guangmin Zheng, Jin Wang, Xiaobing Zhou et al.

Chain of thought (CoT) has proven useful for problems requiring complex reasoning. Many of these problems are both textual and multimodal. Given the inputs in different modalities, a model generates a rationale and then uses it to answer a question. Because of the hallucination issue, the generated soft negative rationales with high textual quality but illogical semantics do not always help improve answer accuracy. This study proposes a rationale generation method using soft negative sampling (SNSE-CoT) to mitigate hallucinations in multimodal CoT. Five methods were applied to generate soft negative samples that shared highly similar text but had different semantics from the original. Bidirectional margin loss (BML) was applied to introduce them into the traditional contrastive learning framework that involves only positive and negative samples. Extensive experiments on the ScienceQA dataset demonstrated the effectiveness of the proposed method. Code and data are released at https://github.com/zgMin/SNSE-CoT.

Jin Wang, Chenghui Lv, Xian Li et al.

Recently, the rapid development of AIGC has significantly boosted the diversities of fake media spread in the Internet, posing unprecedented threats to social security, politics, law, and etc. To detect the ever-increasingly diverse malicious fake media in the new era of AIGC, recent studies have proposed to exploit Large Vision Language Models (LVLMs) to design robust forgery detectors due to their impressive performance on a wide range of multimodal tasks. However, it still lacks a comprehensive benchmark designed to comprehensively assess LVLMs' discerning capabilities on forgery media. To fill this gap, we present Forensics-Bench, a new forgery detection evaluation benchmark suite to assess LVLMs across massive forgery detection tasks, requiring comprehensive recognition, location and reasoning capabilities on diverse forgeries. Forensics-Bench comprises 63,292 meticulously curated multi-choice visual questions, covering 112 unique forgery detection types from 5 perspectives: forgery semantics, forgery modalities, forgery tasks, forgery types and forgery models. We conduct thorough evaluations on 22 open-sourced LVLMs and 3 proprietary models GPT-4o, Gemini 1.5 Pro, and Claude 3.5 Sonnet, highlighting the significant challenges of comprehensive forgery detection posed by Forensics-Bench. We anticipate that Forensics-Bench will motivate the community to advance the frontier of LVLMs, striving for all-around forgery detectors in the era of AIGC. The deliverables will be updated at https://Forensics-Bench.github.io/.

Yibo Zhang, Liang Lin, Kaiwen Luo et al.

While Audio Large Models (ALMs) have achieved remarkable proficiency, their robustness remains brittle in real-world deployment. Existing evaluations largely rely on synthetic Gaussian noise or simplistic single-source interference, failing to capture the intricate, multi-layered acoustic dynamics -- or ``Acoustic Ecology'' -- that characterize authentic physical environments. To bridge this ecological gap, we introduce \textbf{RSA-Bench}, a comprehensive robustness benchmark designed to stress-test ALLMs through high-fidelity auditory scene simulations. Unlike traditional methods, we construct evaluation samples by naturally superimposing diverse environmental soundscapes -- spanning \textit{Pasture}, \textit{Extreme Weather}, \textit{Classroom}, and \textit{Outdoors} -- onto clean speech signals across a spectrum of interference intensities. By evaluating models on six core tasks ranging from fundamental perception to complex reasoning, our study unveils three macro-level insights: \textbf{(I) The Perception-Cognition Gap:} Models maintain relative resilience in low-level recognition but suffer a \textbf{functional collapse} in high-order reasoning tasks under stress; \textbf{(II) Scenario Sensitivity:} ``Vocal-like'' interference (e.g., background laughter) proves significantly more destructive than mechanical noise, challenging the model's auditory attention mechanisms; and \textbf{(III) The Denoising Paradox:} Standard speech enhancement often exacerbates performance degradation, as ALLMs prove highly sensitive to the semantic distortions introduced by denoising artifacts.

Dachun Kai, Yueyi Zhang, Jin Wang et al.

In this paper, we tackle the task of blurry video super-resolution (BVSR), aiming to generate high-resolution (HR) videos from low-resolution (LR) and blurry inputs. Current BVSR methods often fail to restore sharp details at high resolutions, resulting in noticeable artifacts and jitter due to insufficient motion information for deconvolution and the lack of high-frequency details in LR frames. To address these challenges, we introduce event signals into BVSR and propose a novel event-enhanced network, Ev-DeblurVSR. To effectively fuse information from frames and events for feature deblurring, we introduce a reciprocal feature deblurring module that leverages motion information from intra-frame events to deblur frame features while reciprocally using global scene context from the frames to enhance event features. Furthermore, to enhance temporal consistency, we propose a hybrid deformable alignment module that fully exploits the complementary motion information from inter-frame events and optical flow to improve motion estimation in the deformable alignment process. Extensive evaluations demonstrate that Ev-DeblurVSR establishes a new state-of-the-art performance on both synthetic and real-world datasets. Notably, on real data, our method is +2.59 dB more accurate and 7.28$\times$ faster than the recent best BVSR baseline FMA-Net. Code: https://github.com/DachunKai/Ev-DeblurVSR.

Zibo Xu, Qiang Li, Ke Lu et al.

Medical Visual Question Answering (MedVQA) aims to generate clinically reliable answers conditioned on complex medical images and questions. However, existing methods often overfit to superficial cross-modal correlations, neglecting the intrinsic biases embedded in multimodal medical data. Consequently, models become vulnerable to cross-modal confounding effects, severely hindering their ability to provide trustworthy diagnostic reasoning. To address this limitation, we propose a novel Dual Causal Inference (DCI) framework for MedVQA. To the best of our knowledge, DCI is the first unified architecture that integrates Backdoor Adjustment (BDA) and Instrumental Variable (IV) learning to jointly tackle both observable and unobserved confounders. Specifically, we formulate a Structural Causal Model (SCM) where observable cross-modal biases (e.g., frequent visual and textual co-occurrences) are mitigated via BDA, while unobserved confounders are compensated using an IV learned from a shared latent space. To guarantee the validity of the IV, we design mutual information constraints that maximize its dependence on the fused multimodal representations while minimizing its associations with the unobserved confounders and target answers. Through this dual mechanism, DCI extracts deconfounded representations that capture genuine causal relationships. Extensive experiments on four benchmark datasets, SLAKE, SLAKE-CP, VQA-RAD, and PathVQA, demonstrate that our method consistently outperforms existing approaches, particularly in out-of-distribution (OOD) generalization. Furthermore, qualitative analyses confirm that DCI significantly enhances the interpretability and robustness of cross-modal reasoning by explicitly disentangling true causal effects from spurious cross-modal shortcuts.

Jin Wang, Yongsong Zou, Se-Jung Lim

Recurrent Neural Networks (RNNs) have been widely applied to deal with temporal problems, such as flood forecasting and financial data processing. On the one hand, traditional RNNs models amplify the gradient issue due to the strict time serial dependency, making it difficult to realize a long-term memory function. On the other hand, RNNs cells are highly complex, which will significantly increase computational complexity and cause waste of computational resources during model training. In this paper, an improved Time Feedforward Connections Recurrent Neural Networks (TFC-RNNs) model was first proposed to address the gradient issue. A parallel branch was introduced for the hidden state at time t-2 to be directly transferred to time t without the nonlinear transformation at time t-1. This is effective in improving the long-term dependence of RNNs. Then, a novel cell structure named Single Gate Recurrent Unit (SGRU) was presented. This cell structure can reduce the number of parameters for RNNs cell, consequently reducing the computational complexity. Next, applying SGRU to TFC-RNNs as a new TFC-SGRU model solves the above two difficulties. Finally, the performance of our proposed TFC-SGRU was verified through several experiments in terms of long-term memory and anti-interference capabilities. Experimental results demonstrated that our proposed TFC-SGRU model can capture helpful information with time step 1500 and effectively filter out the noise. The TFC-SGRU model accuracy is better than the LSTM and GRU models regarding language processing ability.

Yunjiang Xu, Lingzhi Li, Jin Wang et al.

Collaborative perception, fusing information from multiple agents, can extend perception range so as to improve perception performance. However, temporal asynchrony in real-world environments, caused by communication delays, clock misalignment, or sampling configuration differences, can lead to information mismatches. If this is not well handled, then the collaborative performance is patchy, and what's worse safety accidents may occur. To tackle this challenge, we propose CoDynTrust, an uncertainty-encoded asynchronous fusion perception framework that is robust to the information mismatches caused by temporal asynchrony. CoDynTrust generates dynamic feature trust modulus (DFTM) for each region of interest by modeling aleatoric and epistemic uncertainty as well as selectively suppressing or retaining single-vehicle features, thereby mitigating information mismatches. We then design a multi-scale fusion module to handle multi-scale feature maps processed by DFTM. Compared to existing works that also consider asynchronous collaborative perception, CoDynTrust combats various low-quality information in temporally asynchronous scenarios and allows uncertainty to be propagated to downstream tasks such as planning and control. Experimental results demonstrate that CoDynTrust significantly reduces performance degradation caused by temporal asynchrony across multiple datasets, achieving state-of-the-art detection performance even with temporal asynchrony. The code is available at https://github.com/CrazyShout/CoDynTrust.