Junjie Li

Anthony Cioppa, Silvio Giancola, Vladimir Somers et al. · pku

The SoccerNet 2023 challenges were the third annual video understanding challenges organized by the SoccerNet team. For this third edition, the challenges were composed of seven vision-based tasks split into three main themes. The first theme, broadcast video understanding, is composed of three high-level tasks related to describing events occurring in the video broadcasts: (1) action spotting, focusing on retrieving all timestamps related to global actions in soccer, (2) ball action spotting, focusing on retrieving all timestamps related to the soccer ball change of state, and (3) dense video captioning, focusing on describing the broadcast with natural language and anchored timestamps. The second theme, field understanding, relates to the single task of (4) camera calibration, focusing on retrieving the intrinsic and extrinsic camera parameters from images. The third and last theme, player understanding, is composed of three low-level tasks related to extracting information about the players: (5) re-identification, focusing on retrieving the same players across multiple views, (6) multiple object tracking, focusing on tracking players and the ball through unedited video streams, and (7) jersey number recognition, focusing on recognizing the jersey number of players from tracklets. Compared to the previous editions of the SoccerNet challenges, tasks (2-3-7) are novel, including new annotations and data, task (4) was enhanced with more data and annotations, and task (6) now focuses on end-to-end approaches. More information on the tasks, challenges, and leaderboards are available on https://www.soccer-net.org. Baselines and development kits can be found on https://github.com/SoccerNet.

Junjie Li, Yichao Yan, Guanshuo Wang et al.

Person search is a challenging task which aims to achieve joint pedestrian detection and person re-identification (ReID). Previous works have made significant advances under fully and weakly supervised settings. However, existing methods ignore the generalization ability of the person search models. In this paper, we take a further step and present Domain Adaptive Person Search (DAPS), which aims to generalize the model from a labeled source domain to the unlabeled target domain. Two major challenges arises under this new setting: one is how to simultaneously solve the domain misalignment issue for both detection and Re-ID tasks, and the other is how to train the ReID subtask without reliable detection results on the target domain. To address these challenges, we propose a strong baseline framework with two dedicated designs. 1) We design a domain alignment module including image-level and task-sensitive instance-level alignments, to minimize the domain discrepancy. 2) We take full advantage of the unlabeled data with a dynamic clustering strategy, and employ pseudo bounding boxes to support ReID and detection training on the target domain. With the above designs, our framework achieves 34.7% in mAP and 80.6% in top-1 on PRW dataset, surpassing the direct transferring baseline by a large margin. Surprisingly, the performance of our unsupervised DAPS model even surpasses some of the fully and weakly supervised methods. The code is available at https://github.com/caposerenity/DAPS.

Junjie Li, Zilei Wang, Yuan Gao et al.

In unsupervised domain adaptive (UDA) semantic segmentation, the distillation based methods are currently dominant in performance. However, the distillation technique requires complicate multi-stage process and many training tricks. In this paper, we propose a simple yet effective method that can achieve competitive performance to the advanced distillation methods. Our core idea is to fully explore the target-domain information from the views of boundaries and features. First, we propose a novel mix-up strategy to generate high-quality target-domain boundaries with ground-truth labels. Different from the source-domain boundaries in previous works, we select the high-confidence target-domain areas and then paste them to the source-domain images. Such a strategy can generate the object boundaries in target domain (edge of target-domain object areas) with the correct labels. Consequently, the boundary information of target domain can be effectively captured by learning on the mixed-up samples. Second, we design a multi-level contrastive loss to improve the representation of target-domain data, including pixel-level and prototype-level contrastive learning. By combining two proposed methods, more discriminative features can be extracted and hard object boundaries can be better addressed for the target domain. The experimental results on two commonly adopted benchmarks (\textit{i.e.}, GTA5 $\rightarrow$ Cityscapes and SYNTHIA $\rightarrow$ Cityscapes) show that our method achieves competitive performance to complicated distillation methods. Notably, for the SYNTHIA$\rightarrow$ Cityscapes scenario, our method achieves the state-of-the-art performance with $57.8\%$ mIoU and $64.6\%$ mIoU on 16 classes and 13 classes. Code is available at https://github.com/ljjcoder/EHTDI.

Guanshuo Wang, Fufu Yu, Junjie Li et al. · pku

Text-based Person Search (TPS), is targeted on retrieving pedestrians to match text descriptions instead of query images. Recent Vision-Language Pre-training (VLP) models can bring transferable knowledge to downstream TPS tasks, resulting in more efficient performance gains. However, existing TPS methods improved by VLP only utilize pre-trained visual encoders, neglecting the corresponding textual representation and breaking the significant modality alignment learned from large-scale pre-training. In this paper, we explore the full utilization of textual potential from VLP in TPS tasks. We build on the proposed VLP-TPS baseline model, which is the first TPS model with both pre-trained modalities. We propose the Multi-Integrity Description Constraints (MIDC) to enhance the robustness of the textual modality by incorporating different components of fine-grained corpus during training. Inspired by the prompt approach for zero-shot classification with VLP models, we propose the Dynamic Attribute Prompt (DAP) to provide a unified corpus of fine-grained attributes as language hints for the image modality. Extensive experiments show that our proposed TPS framework achieves state-of-the-art performance, exceeding the previous best method by a margin.

Silvio Giancola, Anthony Cioppa, Adrien Deliège et al.

The SoccerNet 2022 challenges were the second annual video understanding challenges organized by the SoccerNet team. In 2022, the challenges were composed of 6 vision-based tasks: (1) action spotting, focusing on retrieving action timestamps in long untrimmed videos, (2) replay grounding, focusing on retrieving the live moment of an action shown in a replay, (3) pitch localization, focusing on detecting line and goal part elements, (4) camera calibration, dedicated to retrieving the intrinsic and extrinsic camera parameters, (5) player re-identification, focusing on retrieving the same players across multiple views, and (6) multiple object tracking, focusing on tracking players and the ball through unedited video streams. Compared to last year's challenges, tasks (1-2) had their evaluation metrics redefined to consider tighter temporal accuracies, and tasks (3-6) were novel, including their underlying data and annotations. More information on the tasks, challenges and leaderboards are available on https://www.soccer-net.org. Baselines and development kits are available on https://github.com/SoccerNet.

Junjie Li, Yang Liu, Weiqing Liu et al.

Generative models aim to simulate realistic effects of various actions across different contexts, from text generation to visual effects. Despite significant efforts to build real-world simulators, the application of generative models to virtual worlds, like financial markets, remains under-explored. In financial markets, generative models can simulate complex market effects of participants with various behaviors, enabling interaction under different market conditions, and training strategies without financial risk. This simulation relies on the finest structured data in financial market like orders thus building the finest realistic simulation. We propose Large Market Model (LMM), an order-level generative foundation model, for financial market simulation, akin to language modeling in the digital world. Our financial Market Simulation engine (MarS), powered by LMM, addresses the domain-specific need for realistic, interactive and controllable order generation. Key observations include LMM's strong scalability across data size and model complexity, and MarS's robust and practicable realism in controlled generation with market impact. We showcase MarS as a forecast tool, detection system, analysis platform, and agent training environment, thus demonstrating MarS's "paradigm shift" potential for a variety of financial applications. We release the code of MarS at https://github.com/microsoft/MarS/.

Junyi Peng, Oldřich Plchot, Xiao Song et al.

Modern speaker verification (SV) systems rely on speaker embeddings that are effective but difficult to interpret or query in natural language. Most existing speech-text corpora target controllable synthesis or utterance-level captioning, and provide limited speaker-level supervision for in-the-wild speaker recognition. This paper introduces SpeakerCard-1M, a bilingual speaker-centric resource for evidence-grounded SV, derived from VoxCeleb1/2 and CN-Celeb1/2, where the "-1M" suffix refers to the 1.78M utterance-level captions contained in the release. We adopt a tool-first, LLM-last approach: ten acoustic probes produce field-level evidence, the evidence is aggregated into speaker profiles under a schema that separates relatively stable traits from utterance-level states, and bilingual Speaker Cards are rendered by a constrained LLM that sees only the structured fields. The release includes 56.7K Speaker Card records over 10.2K speakers, 1.78M utterance-level captions, and speaker-ID-disjoint hard-negative triplets. We further define two SV-oriented cross-modal protocols, bidirectional Speaker-Text Retrieval (T2S-R / S2T-R) and Attribute-Conditioned Verification (AC-Verify), and compare a dual-encoder baseline against recent audio language models under a zero-shot forced-choice setting. Joint audio-text training increases VoxCeleb1-O EER by 0.31% absolute over the audio-only baseline. Under a style-symmetric LLM-generated counterfactual protocol, eight recent audio language models (7B-30B+ parameters, both open- and closed-source) score 49-77% on pitch-level AC-Verify under two-way forced choice, compared with 88.66% reached by our dual encoder.

Jiafan Zhuang, Zilei Wang, Junjie Li

Video semantic segmentation aims to generate accurate semantic maps for each video frame. To this end, many works dedicate to integrate diverse information from consecutive frames to enhance the features for prediction, where a feature alignment procedure via estimated optical flow is usually required. However, the optical flow would inevitably suffer from inaccuracy, and then introduce noises in feature fusion and further result in unsatisfactory segmentation results. In this paper, to tackle the misalignment issue, we propose a spatial-temporal fusion (STF) module to model dense pairwise relationships among multi-frame features. Different from previous methods, STF uniformly and adaptively fuses features at different spatial and temporal positions, and avoids error-prone optical flow estimation. Besides, we further exploit feature refinement within a single frame and propose a novel memory-augmented refinement (MAR) module to tackle difficult predictions among semantic boundaries. Specifically, MAR can store the boundary features and prototypes extracted from the training samples, which together form the task-specific memory, and then use them to refine the features during inference. Essentially, MAR can move the hard features closer to the most likely category and thus make them more discriminative. We conduct extensive experiments on Cityscapes and CamVid, and the results show that our proposed methods significantly outperform previous methods and achieves the state-of-the-art performance. Code and pretrained models are available at https://github.com/jfzhuang/ST_Memory.

Junjie Li, Meng Ge, Zexu Pan et al.

Speaker extraction seeks to extract the target speech in a multi-talker scenario given an auxiliary reference. Such reference can be auditory, i.e., a pre-recorded speech, visual, i.e., lip movements, or contextual, i.e., phonetic sequence. References in different modalities provide distinct and complementary information that could be fused to form top-down attention on the target speaker. Previous studies have introduced visual and contextual modalities in a single model. In this paper, we propose a two-stage time-domain visual-contextual speaker extraction network named VCSE, which incorporates visual and self-enrolled contextual cues stage by stage to take full advantage of every modality. In the first stage, we pre-extract a target speech with visual cues and estimate the underlying phonetic sequence. In the second stage, we refine the pre-extracted target speech with the self-enrolled contextual cues. Experimental results on the real-world Lip Reading Sentences 3 (LRS3) database demonstrate that our proposed VCSE network consistently outperforms other state-of-the-art baselines.

Jinggang Chen, Junjie Li, Xiaoyang Qu et al.

Detecting out-of-distribution (OOD) examples is crucial to guarantee the reliability and safety of deep neural networks in real-world settings. In this paper, we offer an innovative perspective on quantifying the disparities between in-distribution (ID) and OOD data -- analyzing the uncertainty that arises when models attempt to explain their predictive decisions. This perspective is motivated by our observation that gradient-based attribution methods encounter challenges in assigning feature importance to OOD data, thereby yielding divergent explanation patterns. Consequently, we investigate how attribution gradients lead to uncertain explanation outcomes and introduce two forms of abnormalities for OOD detection: the zero-deflation abnormality and the channel-wise average abnormality. We then propose GAIA, a simple and effective approach that incorporates Gradient Abnormality Inspection and Aggregation. The effectiveness of GAIA is validated on both commonly utilized (CIFAR) and large-scale (ImageNet-1k) benchmarks. Specifically, GAIA reduces the average FPR95 by 23.10% on CIFAR10 and by 45.41% on CIFAR100 compared to advanced post-hoc methods.

Jinggang Chen, Xiaoyang Qu, Junjie Li et al.

Out-of-distribution (OOD) detection aims at enhancing standard deep neural networks to distinguish anomalous inputs from original training data. Previous progress has introduced various approaches where the in-distribution training data and even several OOD examples are prerequisites. However, due to privacy and security, auxiliary data tends to be impractical in a real-world scenario. In this paper, we propose a data-free method without training on natural data, called Class-Conditional Impressions Reappearing (C2IR), which utilizes image impressions from the fixed model to recover class-conditional feature statistics. Based on that, we introduce Integral Probability Metrics to estimate layer-wise class-conditional deviations and obtain layer weights by Measuring Gradient-based Importance (MGI). The experiments verify the effectiveness of our method and indicate that C2IR outperforms other post-hoc methods and reaches comparable performance to the full access (ID and OOD) detection method, especially in the far-OOD dataset (SVHN).

Junjie Li, Guanshuo Wang, Yichao Yan et al.

Person search is a challenging task that involves detecting and retrieving individuals from a large set of un-cropped scene images. Existing person search applications are mostly trained and deployed in the same-origin scenarios. However, collecting and annotating training samples for each scene is often difficult due to the limitation of resources and the labor cost. Moreover, large-scale intra-domain data for training are generally not legally available for common developers, due to the regulation of privacy and public security. Leveraging easily accessible large-scale User Generated Video Contents (\emph{i.e.} UGC videos) to train person search models can fit the open-world distribution, but still suffering a performance gap from the domain difference to surveillance scenes. In this work, we explore enhancing the out-of-domain generalization capabilities of person search models, and propose a generalizable framework on both feature-level and data-level generalization to facilitate downstream tasks in arbitrary scenarios. Specifically, we focus on learning domain-invariant representations for both detection and ReID by introducing a multi-task prototype-based domain-specific batch normalization, and a channel-wise ID-relevant feature decorrelation strategy. We also identify and address typical sources of noise in open-world training frames, including inaccurate bounding boxes, the omission of identity labels, and the absence of cross-camera data. Our framework achieves promising performance on two challenging person search benchmarks without using any human annotation or samples from the target domain.

Yiwei Guo, Zhihan Li, Junjie Li et al.

We propose a new speech discrete token vocoder, vec2wav 2.0, which advances voice conversion (VC). We use discrete tokens from speech self-supervised models as the content features of source speech, and treat VC as a prompted vocoding task. To amend the loss of speaker timbre in the content tokens, vec2wav 2.0 utilizes the WavLM features to provide strong timbre-dependent information. A novel adaptive Snake activation function is proposed to better incorporate timbre into the waveform reconstruction process. In this way, vec2wav 2.0 learns to alter the speaker timbre appropriately given different reference prompts. Also, no supervised data is required for vec2wav 2.0 to be effectively trained. Experimental results demonstrate that vec2wav 2.0 outperforms all other baselines to a considerable margin in terms of audio quality and speaker similarity in any-to-any VC. Ablation studies verify the effects made by the proposed techniques. Moreover, vec2wav 2.0 achieves competitive cross-lingual VC even only trained on monolingual corpus. Thus, vec2wav 2.0 shows timbre can potentially be manipulated only by speech token vocoders, pushing the frontiers of VC and speech synthesis.

Zengbin Wang, Junjie Li, Saihui Hou et al.

The goal of gait recognition is to extract identity-invariant features of an individual under various gait conditions, e.g., cross-view and cross-clothing. Most gait models strive to implicitly learn the common traits across different gait conditions in a data-driven manner to pull different gait conditions closer for recognition. However, relatively few studies have explicitly explored the inherent relations between different gait conditions. For this purpose, we attempt to establish connections among different gait conditions and propose a new perspective to achieve gait recognition: variations in different gait conditions can be approximately viewed as a combination of geometric transformations. In this case, all we need is to determine the types of geometric transformations and achieve geometric invariance, then identity invariance naturally follows. As an initial attempt, we explore three common geometric transformations (i.e., Reflect, Rotate, and Scale) and design a $\mathcal{R}$eflect-$\mathcal{R}$otate-$\mathcal{S}$cale invariance learning framework, named ${\mathcal{RRS}}$-Gait. Specifically, it first flexibly adjusts the convolution kernel based on the specific geometric transformations to achieve approximate feature equivariance. Then these three equivariant-aware features are respectively fed into a global pooling operation for final invariance-aware learning. Extensive experiments on four popular gait datasets (Gait3D, GREW, CCPG, SUSTech1K) show superior performance across various gait conditions.

Sizhuo Zhou, Xiaosong Jia, Fanrui Zhang et al.

Generative world models for autonomous driving (AD) have become a trending topic. Unlike the widely studied image modality, in this work we explore generative world models for LiDAR data. Existing generation methods for LiDAR data only support single frame generation, while existing prediction approaches require multiple frames of historical input and can only deterministically predict multiple frames at once, lacking interactivity. Both paradigms fail to support long-horizon interactive generation. To this end, we introduce LaGen, which to the best of our knowledge is the first framework capable of frame-by-frame autoregressive generation of long-horizon LiDAR scenes. LaGen is able to take a single-frame LiDAR input as a starting point and effectively utilize bounding box information as conditions to generate high-fidelity 4D scene point clouds. In addition, we introduce a scene decoupling estimation module to enhance the model's interactive generation capability for object-level content, as well as a noise modulation module to mitigate error accumulation during long-horizon generation. We construct a protocol based on nuScenes for evaluating long-horizon LiDAR scene generation. Experimental results comprehensively demonstrate LaGen outperforms state-of-the-art LiDAR generation and prediction models, especially on the later frames.

Haoran Sun, Lixin Liu, Junjie Li et al.

The ability of large language models (LLMs) to follow instructions is crucial to real-world applications. Despite recent advances, several studies have highlighted that LLMs struggle when faced with challenging instructions, especially those that include complex constraints, hindering their effectiveness in various tasks. To address this challenge, we introduce Conifer, a novel instruction tuning dataset, designed to enhance LLMs to follow multi-level instructions with complex constraints. Utilizing GPT-4, we curate the dataset by a series of LLM-driven refinement processes to ensure high quality. We also propose a progressive learning scheme that emphasizes an easy-to-hard progression, and learning from process feedback. Models trained with Conifer exhibit remarkable improvements in instruction-following abilities, especially for instructions with complex constraints. On several instruction-following benchmarks, our 7B model outperforms the state-of-the-art open-source 7B models, even exceeds the performance of models 10 times larger on certain metrics. All the code and Conifer dataset are available at https://www.github.com/ConiferLM/Conifer.

Qinying Liu, Zilei Wang, Shenghai Rong et al.

Weakly-supervised temporal action localization aims to localize action instances in videos with only video-level action labels. Existing methods mainly embrace a localization-by-classification pipeline that optimizes the snippet-level prediction with a video classification loss. However, this formulation suffers from the discrepancy between classification and detection, resulting in inaccurate separation of foreground and background (F\&B) snippets. To alleviate this problem, we propose to explore the underlying structure among the snippets by resorting to unsupervised snippet clustering, rather than heavily relying on the video classification loss. Specifically, we propose a novel clustering-based F\&B separation algorithm. It comprises two core components: a snippet clustering component that groups the snippets into multiple latent clusters and a cluster classification component that further classifies the cluster as foreground or background. As there are no ground-truth labels to train these two components, we introduce a unified self-labeling mechanism based on optimal transport to produce high-quality pseudo-labels that match several plausible prior distributions. This ensures that the cluster assignments of the snippets can be accurately associated with their F\&B labels, thereby boosting the F\&B separation. We evaluate our method on three benchmarks: THUMOS14, ActivityNet v1.2 and v1.3. Our method achieves promising performance on all three benchmarks while being significantly more lightweight than previous methods. Code is available at https://github.com/Qinying-Liu/CASE

Junjie Li, Congyang Ou, Haokui Zhang et al.

Recently, diffusion models bring novel insights for Pan-sharpening and notably boost fusion precision. However, most existing models perform diffusion in the pixel space and train distinct models for different multispectral (MS) imagery, suffering from high latency and sensor-specific limitations. In this paper, we present SALAD-Pan, a sensor-agnostic latent space diffusion method for efficient pansharpening. Specifically, SALAD-Pan trains a band-wise single-channel VAE to encode high-resolution multispectral (HRMS) into compact latent representations, supporting MS images with various channel counts and establishing a basis for acceleration. Then spectral physical properties, along with PAN and MS images, are injected into the diffusion backbone through unidirectional and bidirectional interactive control structures respectively, achieving high-precision fusion in the diffusion process. Finally, a lightweight cross-spectral attention module is added to the central layer of diffusion model, reinforcing spectral connections to boost spectral consistency and further elevate fusion precision. Experimental results on GaoFen-2, QuickBird, and WorldView-3 demonstrate that SALAD-Pan outperforms state-of-the-art diffusion-based methods across all three datasets, attains a 2-3x inference speedup, and exhibits robust zero-shot (cross-sensor) capability.

Junjie Li, Jietong Zhao, Yanqing Su et al.

High order structures (cavities and cliques) of the gene network of influenza A virus reveal tight associations among viruses during evolution and are key signals that indicate viral cross-species infection and cause pandemics. As indicators for sensing the dynamic changes of viral genes, these higher order structures have been the focus of attention in the field of virology. However, the size of the viral gene network is usually huge, and searching these structures in the networks introduces unacceptable delay. To mitigate this issue, in this paper, we propose a simple-yet-effective model named HyperSearch based on deep learning to search cavities in a computable complex network for influenza virus genetics. Extensive experiments conducted on a public influenza virus dataset demonstrate the effectiveness of HyperSearch over other advanced deep-learning methods without any elaborated model crafting. Moreover, HyperSearch can finish the search works in minutes while 0-1 programming takes days. Since the proposed method is simple and easy to be transferred to other complex networks, HyperSearch has the potential to facilitate the monitoring of dynamic changes in viral genes and help humans keep up with the pace of virus mutations.

Zishuo Lan, Junjie Li, Lei Wang et al.

Autoregressive learning of time-stepping operators offers an effective approach to data-driven PDE simulation on grids. For conservation laws, however, long-horizon rollouts are often destabilized when learned updates violate global conservation and, in many applications, additional state bounds such as nonnegative mass and densities or concentrations constrained to [0,1]. Enforcing these coupled constraints via direct next-state regression remains difficult. We introduce a framework for learning conservative transport operators on regular grids, inspired by lattice Boltzmann-style discrete-velocity transport representations. Instead of predicting the next state, the model outputs local transport operators that update cells through neighborhood exchanges, guaranteeing discrete conservation by construction. For bounded quantities, we parameterize transport within a capacity-constrained feasible set, enforcing bounds structurally rather than by post-hoc clipping. We validate FluxNet on 1D convection-diffusion, 2D shallow water equations, 1D traffic flow, and 2D spinodal decomposition. Experiments on shallow-water equations and traffic flow show improved rollout stability and physical consistency over strong baselines. On phase-field spinodal decomposition, the method enables large time-steps with long-range transport, accelerating simulation while preserving microstructure evolution in both pointwise and statistical measures.

Pingyi Fan, Anbai Jiang, Shuwei Zhang et al.

With the rapid deployment of SCADA systems, how to effectively analyze industrial signals and detect abnormal states is an urgent need for the industry. Due to the significant heterogeneity of these signals, which we summarize as the M5 problem, previous works only focus on small sub-problems and employ specialized models, failing to utilize the synergies between modalities and the powerful scaling law. However, we argue that the M5 signals can be modeled in a unified manner due to the intrinsic similarity. As a result, we propose FISHER, a Foundation model for multi-modal Industrial Signal compreHEnsive Representation. To support arbitrary sampling rates, FISHER considers the increment of sampling rate as the concatenation of sub-band information. Specifically, FISHER takes the STFT sub-band as the modeling unit and adopts a teacher student SSL framework for pre-training. We also develop the RMIS benchmark, which evaluates the representations of M5 industrial signals on multiple health management tasks. Compared with top SSL models, FISHER showcases versatile and outstanding capabilities with a general performance gain up to 5.03%, along with much more efficient scaling curves. We also investigate the scaling law on downstream tasks and derive potential avenues for future works. FISHER is now open-sourced on https://github.com/jianganbai/FISHER

Junjie Li, Ziao Wang, NingXuan Ma et al.

Existing reasoning data curation pipelines score whole samples, treating every intermediate step as equally valuable. In reality, steps within a trace contribute very unevenly, and selecting reasoning data well requires assessing them individually. We present GRACE, a gradient-aligned curation method that views each reasoning trace as a sequence of optimization events and scores every step by two complementary signals: its alignment with the answer-oriented gradient direction, and its consistency with the preceding reasoning trajectory. Step-level scores are aggregated into a sample-level value for subset selection, using only the model's internal optimization signals and no external reward models or step annotations. To make this scalable, GRACE introduces a representation-level gradient proxy that estimates step-level alignment from token-level upstream signals in a single forward pass. Post-training Qwen3-VL-2B-Instruct on MMathCoT-1M, GRACE reaches 108.8% of the full-data performance with 20% of the data and retains 100.2% with only 5%, with subsets that transfer effectively across model backbones.

Prajwal Ghimire, Junjie Li, Liu Yaou et al.

Background: Radiogenomics allows identification of radiological biomarkers for genomic phenotypes. In glioblastoma, these biomarkers could potentially complement patient stratification strategies. We aim to develop and analytically validate radiological biomarkers that capture immune cell signatures within IDH-wildtype glioblastoma microenvironment using radiogenomic analysis. Methods: This was a retrospective multicenter study using curated open-access anonymized imaging and genomic data from TCGA-GBM, CPTAC, IvyGAP, REMBRANDT and CGGA datasets. Imaging data consisted of MRI-based radiomic features extracted from necrotic core, enhancing and edema regions of deep learning-based auto-segmented tumors. Radiomic feature selections were performed using nested cross-validated LASSO. Support vector machine and ensemble models were trained using seventeen immune and cell-specific score labels extracted from deconvoluted transcriptomic data using pan-cancer and glioblastoma immune signature matrices as reference standards. Seventeen classifier models trained in three cross-cohort strategies were validated on three held-out datasets assessing stability and generalizability. Results: One-hundred-and-seventy-six patients were included in the study. The immune-related radiomic signatures obtained after feature selection were shape, first order and higher order radiomic features. Models predicting macrophage subtype immune signature showed stable mean performance on balanced accuracy (0.67) and precision (0.89) metrics for three independent holdout datasets with ensemble model outperforming support vector machine model. Conclusion: Radiogenomic models non-invasively predicted the macrophage subtype M0 immune signature in IDH-wildtype glioblastoma. These biomarkers have the potential to stratify patients for immunotherapy within prospective glioblastoma clinical trials.

Xingfeng Li, Xiaohan Shi, Junjie Li et al.

This study introduces EM2LDL, a novel multilingual speech corpus designed to advance mixed emotion recognition through label distribution learning. Addressing the limitations of predominantly monolingual and single-label emotion corpora \textcolor{black}{that restrict linguistic diversity, are unable to model mixed emotions, and lack ecological validity}, EM2LDL comprises expressive utterances in English, Mandarin, and Cantonese, capturing the intra-utterance code-switching prevalent in multilingual regions like Hong Kong and Macao. The corpus integrates spontaneous emotional expressions from online platforms, annotated with fine-grained emotion distributions across 32 categories. Experimental baselines using self-supervised learning models demonstrate robust performance in speaker-independent gender-, age-, and personality-based evaluations, with HuBERT-large-EN achieving optimal results. By incorporating linguistic diversity and ecological validity, EM2LDL enables the exploration of complex emotional dynamics in multilingual settings. This work provides a versatile testbed for developing adaptive, empathetic systems for applications in affective computing, including mental health monitoring and cross-cultural communication. The dataset, annotations, and baseline codes are publicly available at https://github.com/xingfengli/EM2LDL.

Guowei Zhong, Junjie Li, Huaiyu Zhu et al.

In recent years, Multimodal Emotion Recognition (MER) has made substantial progress. Nevertheless, most existing approaches neglect the semantic inconsistencies that may arise across modalities, such as conflicting emotional cues between text and visual inputs. Besides, current methods are often dominated by the text modality due to its strong representational capacity, which can compromise recognition accuracy. To address these challenges, we propose a model termed Calibrated Multimodal Consensus (CMC). CMC introduces a Pseudo Label Generation Module (PLGM) to produce pseudo unimodal labels, enabling unimodal pretraining in a self-supervised fashion. It then employs a Parameter-free Fusion Module (PFM) and a Multimodal Consensus Router (MCR) for multimodal finetuning, thereby mitigating text dominance and guiding the fusion process toward a more reliable consensus. Experimental results demonstrate that CMC achieves performance on par with or superior to state-of-the-art methods across four datasets, CH-SIMS, CH-SIMS v2, CMU-MOSI, and CMU-MOSEI, and exhibits notable advantages in scenarios with semantic inconsistencies on CH-SIMS and CH-SIMS v2. The implementation of this work is publicly accessible at https://github.com/gw-zhong/CMC.

Yang Zhou, Junjie Li, CongYang Ou et al.

Due to its extensive applications, aerial image object detection has long been a hot topic in computer vision. In recent years, advancements in Unmanned Aerial Vehicles (UAV) technology have further propelled this field to new heights, giving rise to a broader range of application requirements. However, traditional UAV aerial object detection methods primarily focus on detecting predefined categories, which significantly limits their applicability. The advent of cross-modal text-image alignment (e.g., CLIP) has overcome this limitation, enabling open-vocabulary object detection (OVOD), which can identify previously unseen objects through natural language descriptions. This breakthrough significantly enhances the intelligence and autonomy of UAVs in aerial scene understanding. This paper presents a comprehensive survey of OVOD in the context of UAV aerial scenes. We begin by aligning the core principles of OVOD with the unique characteristics of UAV vision, setting the stage for a specialized discussion. Building on this foundation, we construct a systematic taxonomy that categorizes existing OVOD methods for aerial imagery and provides a comprehensive overview of the relevant datasets. This structured review enables us to critically dissect the key challenges and open problems at the intersection of these fields. Finally, based on this analysis, we outline promising future research directions and application prospects. This survey aims to provide a clear road map and a valuable reference for both newcomers and seasoned researchers, fostering innovation in this rapidly evolving domain. We keep tracing related works at https://github.com/zhouyang2002/OVOD-in-UVA-imagery

Yixin Zhang, Junjie Li, Zilei Wang

Domain adaptation (DA) aims to transfer knowledge from a label-rich source domain to a related but label-scarce target domain. The conventional DA strategy is to align the feature distributions of the two domains. Recently, increasing researches have focused on self-training or other semi-supervised algorithms to explore the data structure of the target domain. However, the bulk of them depend largely on confident samples in order to build reliable pseudo labels, prototypes or cluster centers. Representing the target data structure in such a way would overlook the huge low-confidence samples, resulting in sub-optimal transferability that is biased towards the samples similar to the source domain. To overcome this issue, we propose a novel contrastive learning method by processing low-confidence samples, which encourages the model to make use of the target data structure through the instance discrimination process. To be specific, we create positive and negative pairs only using low-confidence samples, and then re-represent the original features with the classifier weights rather than directly utilizing them, which can better encode the task-specific semantic information. Furthermore, we combine cross-domain mixup to augment the proposed contrastive loss. Consequently, the domain gap can be well bridged through contrastive learning of intermediate representations across domains. We evaluate the proposed method in both unsupervised and semi-supervised DA settings, and extensive experimental results on benchmarks reveal that our method is effective and achieves state-of-the-art performance. The code can be found in https://github.com/zhyx12/MixLRCo.

Junjie Li, Yixin Zhang, Zilei Wang et al.

Contrastive learning has shown impressive success in enhancing feature discriminability for various visual tasks in a self-supervised manner, but the standard contrastive paradigm (features+$\ell_{2}$ normalization) has limited benefits when applied in domain adaptation. We find that this is mainly because the class weights (weights of the final fully connected layer) are ignored in the domain adaptation optimization process, which makes it difficult for features to cluster around the corresponding class weights. To solve this problem, we propose the \emph{simple but powerful} Probabilistic Contrastive Learning (PCL), which moves beyond the standard paradigm by removing $\ell_{2}$ normalization and replacing the features with probabilities. PCL can guide the probability distribution towards a one-hot configuration, thus minimizing the discrepancy between features and class weights. We conduct extensive experiments to validate the effectiveness of PCL and observe consistent performance gains on five tasks, i.e., Unsupervised/Semi-Supervised Domain Adaptation (UDA/SSDA), Semi-Supervised Learning (SSL), UDA Detection and Semantic Segmentation. Notably, for UDA Semantic Segmentation on SYNTHIA, PCL surpasses the sophisticated CPSL-D by $>\!2\%$ in terms of mean IoU with a much lower training cost (PCL: 1*3090, 5 days v.s. CPSL-D: 4*V100, 11 days). Code is available at https://github.com/ljjcoder/Probabilistic-Contrastive-Learning.

Junjie Li

Static bug detection tools help developers detect code problems. However, it is known that they remain underutilized due to various reasons. Recent advances to incorporate static bug detectors in modern software development workflows can better motivate developers to fix the reported warnings on the fly. In this paper, we study the effectiveness of the state-of-the-art (SOA) solution in tracking warnings by static bug detectors and propose a better solution based on our analysis of the insufficiencies of the SOA solution. In particular, we examined four large-scale open-source systems and crafted a data set of 3,452 static code warnings by two static bug detectors. We manually uncover the ground-truth evolution status of the selected warnings: persistent, resolved, or newly-introduced. Moreover, upon manual analysis, we identified the critical reasons behind the insufficiencies of the SOA matching algorithm. Finally, we propose a better approach to improve the tracking of static warnings over software development history. Our evaluation shows that our proposed approach provides a significant improvement in the precision of the tracking, i.e., from 66.9% to 90.0%.

Junjie Li, Kong Aik Lee

An utterance-level speaker embedding is typically obtained by aggregating a sequence of frame-level representations. However, in real-world scenarios, individual frames encode not only speaker-relevant information but also various nuisance factors. As a result, different frames contribute unequally to the final utterance-level speaker representation for Automatic Speaker Verification systems. To address this issue, we propose to estimate the inherent uncertainty of each frame and assign adaptive weights accordingly, where frames with higher uncertainty receive lower attention. Based on this idea, we present U3-xi, a comprehensive framework designed to produce more reliable and interpretable uncertainty estimates for speaker embeddings. Specifically, we introduce several strategies for uncertainty supervision. First, we propose speaker-level uncertainty supervision via a Stochastic Variance Loss, where the distance between an utterance embedding and its corresponding speaker centroid serves as a pseudo ground truth for uncertainty learning. Second, we incorporate global-level uncertainty supervision by injecting the predicted uncertainty into the sof tmax scale during training. This adaptive scaling mechanism adjusts the sharpness of the decision boundary according to sample difficulty, providing global guidance. Third, we redesign the uncertainty estimation module by integrating a Transformer encoder with multi-view self-attention, enabling the model to capture rich local and long-range temporal dependencies. Comprehensive experiments demonstrate that U3-xi is model-agnostic and can be seamlessly applied to various speaker encoders. In particular, when applied to ECAPA-TDNN, it achieves 21.1% and 15.57% relative improvements on the VoxCeleb1 test sets in terms of EER and minDCF, respectively.

Junjie Li, Jiong Lou, Jie Li

Efficient long-context inference in Large Language Models (LLMs) is severely constrained by the Key-Value (KV) cache memory wall, yet existing pruning methods force a choice between low-latency heuristics that sacrifice precision and high-precision reconstruction methods that incur prohibitive prefilling overhead. To bridge this scoring-cost--accuracy gap, we propose ProxyKV, a cross-model proxy pruning framework that offloads importance scoring to a lightweight intra-family Small-Model Proxy executed asynchronously to the Large-Model Target. To bridge the architectural gap between heterogeneous models, we design the HybridAxialMapper, which disentangles temporal feature extraction from cross-head alignment, together with a Multi-Granularity Hybrid Loss that shifts the learning objective from rigid regression to relative ranking consistency. Across the Llama-3.1, Qwen-2.5, and Qwen-3 families spanning targets from 7B up to 32B parameters on LongBench, SCBench, and RULER, ProxyKV matches KVZip on aggregate (recovering $\sim$$98.7\%$ of its mean accuracy) while delivering up to a $3.21\times$ prefilling speedup on Llama-3.1-8B (dual-GPU; $\sim$$1.5\times$ shared single-GPU) and sustaining the speedup at contexts up to 170k tokens on Qwen-2.5-7B.

Junjie Li, Kang Li, Dewei Han et al.

AI and robotics technologies have witnessed remarkable advancements in the past decade, revolutionizing work patterns and opportunities in various domains. The application of these technologies has propelled society towards an era of symbiosis between humans and machines. To facilitate efficient communication between humans and intelligent robots, we propose the "Avatar" system, an immersive low-latency panoramic human-robot interaction platform. We have designed and tested a prototype of a rugged mobile platform integrated with edge computing units, panoramic video capture devices, power batteries, robot arms, and network communication equipment. Under favorable network conditions, we achieved a low-latency high-definition panoramic visual experience with a delay of 357ms. Operators can utilize VR headsets and controllers for real-time immersive control of robots and devices. The system enables remote control over vast physical distances, spanning campuses, provinces, countries, and even continents (New York to Shenzhen). Additionally, the system incorporates visual SLAM technology for map and trajectory recording, providing autonomous navigation capabilities. We believe that this intuitive system platform can enhance efficiency and situational experience in human-robot collaboration, and with further advancements in related technologies, it will become a versatile tool for efficient and symbiotic cooperation between AI and humans.

Xumeng Wen, Zihan Liu, Shun Zheng et al.

Recent advancements in long chain-of-thought (CoT) reasoning, particularly through the Group Relative Policy Optimization algorithm used by DeepSeek-R1, have led to significant interest in the potential of Reinforcement Learning with Verifiable Rewards (RLVR) for Large Language Models (LLMs). While RLVR promises to improve reasoning by allowing models to learn from free exploration, there remains debate over whether it truly enhances reasoning abilities or simply boosts sampling efficiency. This paper systematically investigates the impact of RLVR on LLM reasoning. We revisit Pass@K experiments and demonstrate that RLVR can extend the reasoning boundary for both mathematical and coding tasks. This is supported by our introduction of a novel evaluation metric, CoT-Pass@K, which captures reasoning success by accounting for both the final answer and intermediate reasoning steps. Furthermore, we present a theoretical framework explaining RLVR's incentive mechanism, demonstrating how it can encourage correct reasoning even when rewards are based solely on answer correctness. Our analysis of RLVR's training dynamics reveals that it incentivizes correct reasoning early in the process, with substantial improvements in reasoning quality confirmed through extensive evaluations. These findings provide strong evidence of RLVR's potential to enhance LLM reasoning, offering valuable insights into its mechanisms and performance improvements.

Li Zhou, Hao Jiang, Junjie Li et al.

Achieving precise and controllable emotional expression is crucial for producing natural and context-appropriate speech in text-to-speech (TTS) synthesis. However, many emotion-aware TTS systems, including large language model (LLM)-based designs, rely on scaling fixed emotion embeddings or external guidance, limiting their ability to model emotion-specific latent characteristics. To address this gap, we present EmoShift, a lightweight activation-steering framework incorporating a EmoSteer layer, which learns a steering vector for each target emotion in the output embedding space to capture its latent offset and maintain stable, appropriate expression across utterances and categories. With only 10M trainable parameters,less than 1/30 of full fine-tuning, EmoShift outperforms zero-shot and fully fine-tuned baselines in objective and subjective evaluations, enhancing emotional expressiveness while preserving naturalness and speaker similarity. Further analysis confirms the proposed EmoSteer layer's effectiveness and reveals its potential for controllable emotional intensity in speech synthesis.

Junjie Li, Yiwei Guo, Xie Chen et al.

Zero-shot voice conversion (VC) aims to transfer the source speaker timbre to arbitrary unseen target speaker timbre, while keeping the linguistic content unchanged. Although the voice of generated speech can be controlled by providing the speaker embedding of the target speaker, the speaker similarity still lags behind the ground truth recordings. In this paper, we propose SEF-VC, a speaker embedding free voice conversion model, which is designed to learn and incorporate speaker timbre from reference speech via a powerful position-agnostic cross-attention mechanism, and then reconstruct waveform from HuBERT semantic tokens in a non-autoregressive manner. The concise design of SEF-VC enhances its training stability and voice conversion performance. Objective and subjective evaluations demonstrate the superiority of SEF-VC to generate high-quality speech with better similarity to target reference than strong zero-shot VC baselines, even for very short reference speeches.

Jiutian Zeng, Junjie Li, Chengwei Dai et al.

We present XekRung, a frontier large language model for cybersecurity, designed to provide comprehensive security capabilities. To achieve this, we develop diverse data synthesis pipelines tailored to the cybersecurity domain, enabling the scalable construction of high-quality training data and providing a strong foundation for cybersecurity knowledge and understanding. Building on this foundation, we establish a complete training pipeline spanning continued pre-training (CPT), supervised fine-tuning (SFT), and reinforcement learning (RL) to further extend the model's capabilities. We further introduce a multi-dimensional evaluation system to guide the iterative improvement of both domain-specific and general-purpose abilities. Extensive experiments demonstrate that XekRung achieves state-of-the-art performance on cybersecurity-specific benchmarks among models of the same scale, while maintaining strong performance on general benchmarks.

Junjie Li, Xinrui Guo, Yuhao Wu et al.

What happens when a storyteller forgets its own story? Large Language Models (LLMs) can now generate narratives spanning tens of thousands of words, but they often fail to maintain consistency throughout. When generating long-form narratives, these models can contradict their own established facts, character traits, and world rules. Existing story generation benchmarks focus mainly on plot quality and fluency, leaving consistency errors largely unexplored. To address this gap, we present ConStory-Bench, a benchmark designed to evaluate narrative consistency in long-form story generation. It contains 2,000 prompts across four task scenarios and defines a taxonomy of five error categories with 19 fine-grained subtypes. We also develop ConStory-Checker, an automated pipeline that detects contradictions and grounds each judgment in explicit textual evidence. Evaluating a range of LLMs through five research questions, we find that consistency errors show clear tendencies: they are most common in factual and temporal dimensions, tend to appear around the middle of narratives, occur in text segments with higher token-level entropy, and certain error types tend to co-occur. These findings can inform future efforts to improve consistency in long-form narrative generation. Our project page is available at https://picrew.github.io/constory-bench.github.io/.

Ziqi Jia, Junjie Li, Xiaoyang Qu et al.

Multi-agent systems (MAS) have shown great potential in executing complex tasks, but coordination and safety remain significant challenges. Multi-Agent Reinforcement Learning (MARL) offers a promising framework for agent collaboration, but it faces difficulties in handling complex tasks and designing reward functions. The introduction of Large Language Models (LLMs) has brought stronger reasoning and cognitive abilities to MAS, but existing LLM-based systems struggle to respond quickly and accurately in dynamic environments. To address these challenges, we propose LLM-based Graph Collaboration MARL (LGC-MARL), a framework that efficiently combines LLMs and MARL. This framework decomposes complex tasks into executable subtasks and achieves efficient collaboration among multiple agents through graph-based coordination. Specifically, LGC-MARL consists of two main components: an LLM planner and a graph-based collaboration meta policy. The LLM planner transforms complex task instructions into a series of executable subtasks, evaluates the rationality of these subtasks using a critic model, and generates an action dependency graph. The graph-based collaboration meta policy facilitates communication and collaboration among agents based on the action dependency graph, and adapts to new task environments through meta-learning. Experimental results on the AI2-THOR simulation platform demonstrate the superior performance and scalability of LGC-MARL in completing various complex tasks.

Xizhe Xue, Yang Zhou, Dawei Yan et al.

Recently, the remarkable success of large language models (LLMs) has achieved a profound impact on the field of artificial intelligence. Numerous advanced works based on LLMs have been proposed and applied in various scenarios. Among them, video language models (VidLMs) are particularly widely used. However, existing works primarily focus on terrestrial scenarios, overlooking the highly demanding application needs of underwater observation. To overcome this gap, we introduce UVLM, an under water observation benchmark which is build through a collaborative approach combining human expertise and AI models. To ensure data quality, we have conducted in-depth considerations from multiple perspectives. First, to address the unique challenges of underwater environments, we selected videos that represent typical underwater challenges including light variations, water turbidity, and diverse viewing angles to construct the dataset. Second, to ensure data diversity, the dataset covers a wide range of frame rates, resolutions, 419 classes of marine animals, and various static plants and terrains. Next, for task diversity, we adopted a structured design where observation targets are categorized into two major classes: biological and environmental. Each category includes content observation and change/action observation, totaling 20 distinct task types. Finally, we designed several challenging evaluation metrics to enable quantitative comparison and analysis of different methods. Experiments on two representative VidLMs demonstrate that fine-tuning VidLMs on UVLM significantly improves underwater world understanding while also showing potential for slight improvements on existing in-air VidLM benchmarks, such as VideoMME and Perception text. The dataset and prompt engineering will be released publicly.

Jinfan Liu, Yichao Yan, Junjie Li et al.

Video anomaly detection (VAD) is a challenging task aiming to recognize anomalies in video frames, and existing large-scale VAD researches primarily focus on road traffic and human activity scenes. In industrial scenes, there are often a variety of unpredictable anomalies, and the VAD method can play a significant role in these scenarios. However, there is a lack of applicable datasets and methods specifically tailored for industrial production scenarios due to concerns regarding privacy and security. To bridge this gap, we propose a new dataset, IPAD, specifically designed for VAD in industrial scenarios. The industrial processes in our dataset are chosen through on-site factory research and discussions with engineers. This dataset covers 16 different industrial devices and contains over 6 hours of both synthetic and real-world video footage. Moreover, we annotate the key feature of the industrial process, ie, periodicity. Based on the proposed dataset, we introduce a period memory module and a sliding window inspection mechanism to effectively investigate the periodic information in a basic reconstruction model. Our framework leverages LoRA adapter to explore the effective migration of pretrained models, which are initially trained using synthetic data, into real-world scenarios. Our proposed dataset and method will fill the gap in the field of industrial video anomaly detection and drive the process of video understanding tasks as well as smart factory deployment.

Junjie Li, Guanshuo Wang, Fufu Yu et al.

Clothes-changing person re-identification (CC-ReID) aims to retrieve images of the same person wearing different outfits. Mainstream researches focus on designing advanced model structures and strategies to capture identity information independent of clothing. However, the same-clothes discrimination as the standard ReID learning objective in CC-ReID is persistently ignored in previous researches. In this study, we dive into the relationship between standard and clothes-changing~(CC) learning objectives, and bring the inner conflicts between these two objectives to the fore. We try to magnify the proportion of CC training pairs by supplementing high-fidelity clothes-varying synthesis, produced by our proposed Clothes-Changing Diffusion model. By incorporating the synthetic images into CC-ReID model training, we observe a significant improvement under CC protocol. However, such improvement sacrifices the performance under the standard protocol, caused by the inner conflict between standard and CC. For conflict mitigation, we decouple these objectives and re-formulate CC-ReID learning as a multi-objective optimization (MOO) problem. By effectively regularizing the gradient curvature across multiple objectives and introducing preference restrictions, our MOO solution surpasses the single-task training paradigm. Our framework is model-agnostic, and demonstrates superior performance under both CC and standard ReID protocols.

Hong Zhou, Rui Zhang, Peifeng Lai et al.

Nowadays, Vision Transformer (ViT) is widely utilized in various computer vision tasks, owing to its unique self-attention mechanism. However, the model architecture of ViT is complex and often challenging to comprehend, leading to a steep learning curve. ViT developers and users frequently encounter difficulties in interpreting its inner workings. Therefore, a visualization system is needed to assist ViT users in understanding its functionality. This paper introduces EL-VIT, an interactive visual analytics system designed to probe the Vision Transformer and facilitate a better understanding of its operations. The system consists of four layers of visualization views. The first three layers include model overview, knowledge background graph, and model detail view. These three layers elucidate the operation process of ViT from three perspectives: the overall model architecture, detailed explanation, and mathematical operations, enabling users to understand the underlying principles and the transition process between layers. The fourth interpretation view helps ViT users and experts gain a deeper understanding by calculating the cosine similarity between patches. Our two usage scenarios demonstrate the effectiveness and usability of EL-VIT in helping ViT users understand the working mechanism of ViT.

Jing Peng, Chenghao Wang, Yi Yang et al.

Speech LLM post-training increasingly relies on efficient cross-modal alignment and robust low-resource adaptation, yet collecting large-scale audio-text pairs remains costly. Text-only alignment methods such as TASU reduce this burden by simulating CTC posteriors from transcripts, but they provide limited control over uncertainty and error rate, making curriculum design largely heuristic. We propose \textbf{TASU2}, a controllable CTC simulation framework that simulates CTC posterior distributions under a specified WER range, producing text-derived supervision that better matches the acoustic decoding interface. This enables principled post-training curricula that smoothly vary supervision difficulty without TTS. Across multiple source-to-target adaptation settings, TASU2 improves in-domain and out-of-domain recognition over TASU, and consistently outperforms strong baselines including text-only fine-tuning and TTS-based augmentation, while mitigating source-domain performance degradation.

Junjie Li, Nan Zhang, Xiaoyang Qu et al.

Object Navigation (ObjectNav) is a fundamental task in embodied artificial intelligence. Although significant progress has been made in semantic map construction and target direction prediction in current research, redundant exploration and exploration failures remain inevitable. A critical but underexplored direction is the timely termination of exploration to overcome these challenges. We observe a diminishing marginal effect between exploration steps and exploration rates and analyze the cost-benefit relationship of exploration. Inspired by this, we propose RATE-Nav, a Region-Aware Termination-Enhanced method. It includes a geometric predictive region segmentation algorithm and region-Based exploration estimation algorithm for exploration rate calculation. By leveraging the visual question answering capabilities of visual language models (VLMs) and exploration rates enables efficient termination.RATE-Nav achieves a success rate of 67.8% and an SPL of 31.3% on the HM3D dataset. And on the more challenging MP3D dataset, RATE-Nav shows approximately 10% improvement over previous zero-shot methods.

Junjie Li, Jiawei Wang, Miyu Li et al.

Depth estimation plays a great potential role in obstacle avoidance and navigation for further Mars exploration missions. Compared to traditional stereo matching, learning-based stereo depth estimation provides a data-driven approach to infer dense and precise depth maps from stereo image pairs. However, these methods always suffer performance degradation in environments with sparse textures and lacking geometric constraints, such as the unstructured terrain of Mars. To address these challenges, we propose M3Depth, a depth estimation model tailored for Mars rovers. Considering the sparse and smooth texture of Martian terrain, which is primarily composed of low-frequency features, our model incorporates a convolutional kernel based on wavelet transform that effectively captures low-frequency response and expands the receptive field. Additionally, we introduce a consistency loss that explicitly models the complementary relationship between depth map and surface normal map, utilizing the surface normal as a geometric constraint to enhance the accuracy of depth estimation. Besides, a pixel-wise refinement module with mutual boosting mechanism is designed to iteratively refine both depth and surface normal predictions. Experimental results on synthetic Mars datasets with depth annotations show that M3Depth achieves a 16% improvement in depth estimation accuracy compared to other state-of-the-art methods in depth estimation. Furthermore, the model demonstrates strong applicability in real-world Martian scenarios, offering a promising solution for future Mars exploration missions.

Junjie Li, Ke Zhang, Shuai Wang et al.

Audio-visual Target Speaker Extraction (AV-TSE) aims to isolate the speech of a specific target speaker from an audio mixture using time-synchronized visual cues. In real-world scenarios, visual cues are not always available due to various impairments, which undermines the stability of AV-TSE. Despite this challenge, humans can maintain attentional momentum over time, even when the target speaker is not visible. In this paper, we introduce the Momentum Multi-modal target Speaker Extraction (MoMuSE), which retains a speaker identity momentum in memory, enabling the model to continuously track the target speaker. Designed for real-time inference, MoMuSE extracts the current speech window with guidance from both visual cues and dynamically updated speaker momentum. Experimental results demonstrate that MoMuSE exhibits significant improvement, particularly in scenarios with severe impairment of visual cues.

Junjie Li, Jianghong Ma, Xiaofeng Zhang et al.

Multimodal Large Language Models have advanced AI in applications like text-to-video generation and visual question answering. These models rely on visual encoders to convert non-text data into vectors, but current encoders either lack semantic alignment or overlook non-salient objects. We propose the Guiding Visual Encoder to Perceive Overlooked Information (GiVE) approach. GiVE enhances visual representation with an Attention-Guided Adapter (AG-Adapter) module and an Object-focused Visual Semantic Learning module. These incorporate three novel loss terms: Object-focused Image-Text Contrast (OITC) loss, Object-focused Image-Image Contrast (OIIC) loss, and Object-focused Image Discrimination (OID) loss, improving object consideration, retrieval accuracy, and comprehensiveness. Our contributions include dynamic visual focus adjustment, novel loss functions to enhance object retrieval, and the Multi-Object Instruction (MOInst) dataset. Experiments show our approach achieves state-of-the-art performance.

Hang Liu, Junjie Li, Yinzhi Wang et al.

This study explores the use of INT8-based emulation for accelerating traditional FP64-based HPC workloads on modern GPU architectures. Through SCILIB-Accel automatic BLAS offload tool for cache-coherent Unified Memory Architecture, we emulate FP64 matrix multiplications in the LSMS CPU application in the MuST suite without code changes. We find that accuracy depends on both arithmetic precision and the properties of the operator, which can be dealt with through tunable precision emulation. Unlike traditional mixed-precision approaches, this method preserves original algorithms while optimizing hardware utilization. We showcase the potential of improving accuracy and performance at the same time. This work highlights the potential of AI-driven hardware to transform HPC, advocating for adaptive precision strategies in future scientific computing.

Junjie Li, Ziao Wang, Jianghong Ma et al.

Large vision-language models (VLMs) achieve strong benchmark performance, but controlling their behavior through instruction tuning remains difficult. Reducing the budget of instruction tuning dataset often causes regressions, as heuristic strategies treat models as black boxes and overlook the latent capabilities that govern learning. We introduce Capability-Attributed Data Curation (CADC), a framework that shifts curation from task-specific heuristics to intrinsic capability analysis. CADC discovers intrinsic capabilities in an unsupervised manner from gradient-based learning trajectories, attributes training data to these capabilities via influence estimation, and curates capability-aware curricula through balanced selection and staged sequencing. This transforms black-box instruction tuning into a controllable, capability-driven process. With as little as 5% of the original data, CADC surpasses full-data training on multimodal benchmarks. These results validate intrinsic capabilities as the fundamental building blocks of model learning and establish CADC as a principle paradigm for instruction data curation.

Duc-Tuan Truong, Tianchi Liu, Junjie Li et al.

In speech deepfake detection (SDD), data augmentation (DA) is commonly used to improve model generalization across varied speech conditions and spoofing attacks. However, during training, the backpropagated gradients from original and augmented inputs may misalign, which can result in conflicting parameter updates. These conflicts could hinder convergence and push the model toward suboptimal solutions, thereby reducing the benefits of DA. To investigate and address this issue, we design a dual-path data-augmented (DPDA) training framework with gradient alignment for SDD. In our framework, each training utterance is processed through two input paths: one using the original speech and the other with its augmented version. This design allows us to compare and align their backpropagated gradient directions to reduce optimization conflicts. Our analysis shows that approximately 25% of training iterations exhibit gradient conflicts between the original inputs and their augmented counterparts when using RawBoost augmentation. By resolving these conflicts with gradient alignment, our method accelerates convergence by reducing the number of training epochs and achieves up to an 18.69% relative reduction in Equal Error Rate on the In-the-Wild dataset compared to the baseline.