Fan Jiang

Fan Jiang, Xingpeng Li

As grid-forming (GFM) battery energy storage systems (BESS) are increasingly deployed to enhance power system inertial response and frequency stability, incorporating their frequency support capabilities into day-ahead energy scheduling (DAES) is essential for achieving both frequency security and operational efficiency. However, accurately determining frequency metrics in grids with coexisting GFM inverters and synchronous generators requires electromagnetic transient (EMT) simulations, which are computationally prohibitive for direct embedding in grid operational optimization models. To bridge the gap between modeling accuracy and computational efficiency, a learning-assisted DAES (LA-DAES) framework is proposed in this work. By leveraging a surrogate model to represent the frequency support dynamics of GFM BESS, the proposed framework ensures frequency security with a reasonable solve time. Comparative results demonstrate that, relative to analytical frequency-constrained DAES, the proposed LA-DAES framework more accurately captures grid frequency metrics and improves the utilization of GFM BESS.

Kedeng Tong, Xin Jin, Yuqing Yang et al.

Focused plenoptic cameras can record spatial and angular information of the light field (LF) simultaneously with higher spatial resolution relative to traditional plenoptic cameras, which facilitate various applications in computer vision. However, the existing plenoptic image compression methods present ineffectiveness to the captured images due to the complex micro-textures generated by the microlens relay imaging and long-distance correlations among the microimages. In this paper, a lossy end-to-end learning architecture is proposed to compress the focused plenoptic images efficiently. First, a data preprocessing scheme is designed according to the imaging principle to remove the sub-aperture image ineffective pixels in the recorded light field and align the microimages to the rectangular grid. Then, the global attention module with large receptive field is proposed to capture the global correlation among the feature maps using pixel-wise vector attention computed in the resampling process. Also, a new image dataset consisting of 1910 focused plenoptic images with content and depth diversity is built to benefit training and testing. Extensive experimental evaluations demonstrate the effectiveness of the proposed approach. It outperforms intra coding of HEVC and VVC by an average of 62.57% and 51.67% bitrate reduction on the 20 preprocessed focused plenoptic images, respectively. Also, it achieves 18.73% bitrate saving and generates perceptually pleasant reconstructions compared to the state-of-the-art end-to-end image compression methods, which benefits the applications of focused plenoptic cameras greatly. The dataset and code are publicly available at https://github.com/VincentChandelier/GACN.

Yufei Xu, Fanxu Meng, Fan Jiang et al.

Token-level sparse attention mechanisms, exemplified by DeepSeek Sparse Attention (DSA), achieve fine-grained key selection by scoring every historical token for each query using a lightweight indexer, and then computing attention only over the selected subset. While the downstream sparse attention scales efficiently, the indexer still scans the entire prefix for every query, introducing an O($L^2$) per-layer bottleneck that becomes prohibitive as context length grows. We propose HISA (Hierarchical Indexed Sparse Attention), a drop-in replacement for the indexer that transforms the search process from a flat token scan into a two-stage hierarchical procedure. First, a block-level coarse filter scores pooled block representatives to prune irrelevant regions. Then, a token-level refinement applies the original indexer only within the remaining candidate blocks. HISA preserves the exact token-level top-k sparsity pattern required by the downstream Sparse MLA operator and requires no additional training. On kernel-level benchmarks, HISA achieves a 2$\times$ speedup at 32K context length and 4$\times$ at 128K. On Needle-in-a-Haystack and LongBench, we directly replace the indexer in DeepSeek-V3.2 with HISA, without any fine-tuning. HISA closely matches the original DSA in quality while significantly outperforming block-sparse baselines. Moreover, the token selection sets produced by HISA and the original DSA exhibit a mean IoU greater than 99%, indicating that the efficiency gains come with virtually no impact on selection fidelity.

Fan Jiang, Qiongkai Xu, Tom Drummond et al.

Neural 'dense' retrieval models are state of the art for many datasets, however these models often exhibit limited domain transfer ability. Existing approaches to adaptation are unwieldy, such as requiring explicit supervision, complex model architectures, or massive external models. We present $\texttt{ABEL}$, a simple but effective unsupervised method to enhance passage retrieval in zero-shot settings. Our technique follows a straightforward loop: a dense retriever learns from supervision signals provided by a reranker, and subsequently, the reranker is updated based on feedback from the improved retriever. By iterating this loop, the two components mutually enhance one another's performance. Experimental results demonstrate that our unsupervised $\texttt{ABEL}$ model outperforms both leading supervised and unsupervised retrievers on the BEIR benchmark. Meanwhile, it exhibits strong adaptation abilities to tasks and domains that were unseen during training. By either fine-tuning $\texttt{ABEL}$ on labelled data or integrating it with existing supervised dense retrievers, we achieve state-of-the-art results.\footnote{Source code is available at \url{https://github.com/Fantabulous-J/BootSwitch}.}

Fan Jiang, Tom Drummond, Trevor Cohn

Although existing neural retrieval models reveal promising results when training data is abundant and the performance keeps improving as training data increases, collecting high-quality annotated data is prohibitively costly. To this end, we introduce a novel noisy self-training framework combined with synthetic queries, showing that neural retrievers can be improved in a self-evolution manner with no reliance on any external models. Experimental results show that our method improves consistently over existing methods on both general-domain (e.g., MS-MARCO) and out-of-domain (i.e., BEIR) retrieval benchmarks. Extra analysis on low-resource settings reveals that our method is data efficient and outperforms competitive baselines, with as little as 30% of labelled training data. Further extending the framework for reranker training demonstrates that the proposed method is general and yields additional gains on tasks of diverse domains.\footnote{Source code is available at \url{https://github.com/Fantabulous-J/Self-Training-DPR}}

Xingwu Sun, Yanfeng Chen, Yiqing Huang et al. · tencent-ai

In this paper, we introduce Hunyuan-Large, which is currently the largest open-source Transformer-based mixture of experts model, with a total of 389 billion parameters and 52 billion activation parameters, capable of handling up to 256K tokens. We conduct a thorough evaluation of Hunyuan-Large's superior performance across various benchmarks including language understanding and generation, logical reasoning, mathematical problem-solving, coding, long-context, and aggregated tasks, where it outperforms LLama3.1-70B and exhibits comparable performance when compared to the significantly larger LLama3.1-405B model. Key practice of Hunyuan-Large include large-scale synthetic data that is orders larger than in previous literature, a mixed expert routing strategy, a key-value cache compression technique, and an expert-specific learning rate strategy. Additionally, we also investigate the scaling laws and learning rate schedule of mixture of experts models, providing valuable insights and guidances for future model development and optimization. The code and checkpoints of Hunyuan-Large are released to facilitate future innovations and applications. Codes: https://github.com/Tencent/Hunyuan-Large Models: https://huggingface.co/tencent/Tencent-Hunyuan-Large

Mingjian Tuo, Fan Jiang, Xingpeng Li et al.

Day-ahead generation scheduling is typically conducted by solv-ing security-constrained unit commitment (SCUC) problem. However, with fast-growing of inverter-based resources, grid inertia has been dramatically reduced, compromising the dy-namic stability system. Traditional SCUC (T-SCUC), without any inertia requirements, may no longer be effective for renewa-bles-dominated grids. To address this, we propose the active linearized sparse neural network-embedded SCUC (ALSNN-SCUC) model, utilizing machine learning (ML) to incorporate system dynamic performance. A multi-output deep neural net-work (DNN) model is trained offline on strategically-selected data samples to accurately predict frequency stability metrics: locational RoCoF and frequency nadir. Structured sparsity and active ReLU linearization are implemented to prune redundant DNN neurons, significantly reducing its size while ensuring pre-diction accuracy even at high sparsity levels. By embedding this ML-based frequency stability predictor into SCUC as con-straints, the proposed ALSNN-SCUC model minimizes its com-putational complexity while ensuring frequency stability follow-ing G-1 contingency. Case studies show that the proposed ALSNN-SCUC can enforce pre-specified frequency requirements without being overly conservative, outperforming five bench-mark models including T-SCUC, two physics-based SCUC, and two ML-based SCUC. The proposed sparsification and active linearization strategies can reduce the DNN-SCUC computing time by over 95% for both IEEE 24-bus and 118-bus systems, demonstrating the effectiveness and scalability of the proposed ALSNN-SCUC model.

Yu Zhao, Fan Jiang, Tianle Liu et al.

Recent advancements in Large Reasoning Models (LRMs), exemplified by DeepSeek-R1, have underscored the potential of scaling inference-time compute through Group Relative Policy Optimization (GRPO). However, GRPO frequently suffers from gradient signal attenuation when encountering problems that are either too trivial or overly complex. In these scenarios, the disappearance of inter-group advantages makes the gradient signal susceptible to noise, thereby jeopardizing convergence stability. While variants like DAPO attempt to rectify gradient vanishing, they do not alleviate the substantial computational overhead incurred by exhaustive rollouts on low-utility samples. In this paper, we propose Difficulty-Estimated Policy Optimization (DEPO), a novel framework designed to optimize the efficiency and robustness of reasoning alignment. DEPO integrates an online Difficulty Estimator that dynamically assesses and filters training data before the rollout phase. This mechanism ensures that computational resources are prioritized for samples with high learning potential. Empirical results demonstrate that DEPO achieves up to a 2x reduction in rollout costs without compromising model performance. Our approach significantly lowers the computational barrier for training high-performance reasoning models, offering a more sustainable path for reasoning scaling. Code and data will be released upon acceptance.

Cunzhi Zhao, Fan Jiang, Xingpeng Li

Recently, neural networks have been widely applied in the power system area. They can be used for better predicting input information and modeling system performance with increased accuracy. In some applications such as battery degradation neural network-based microgrid day-ahead energy scheduling, the input features of the trained learning model are variables to be solved in optimization models that enforce limits on the output of the same learning model. This will create a neural network-embedded optimization problem; the use of nonlinear activation functions in the neural network will make such problems extremely hard to solve if not unsolvable. To address this emerging challenge, this paper investigated different methods for linearizing the nonlinear activation functions with a particular focus on the widely used rectified linear unit (ReLU) function. Four linearization methods tailored for the ReLU activation function are developed, analyzed and compared in this paper. Each method employs a set of linear constraints to replace the ReLU function, effectively linearizing the optimization problem, which can overcome the computational challenges associated with the nonlinearity of the neural network model. These proposed linearization methods provide valuable tools for effectively solving optimization problems that integrate neural network models with ReLU activation functions

Fan Jiang, Trevor Cohn

Syntax has been shown to benefit Coreference Resolution from incorporating long-range dependencies and structured information captured by syntax trees, either in traditional statistical machine learning based systems or recently proposed neural models. However, most leading systems use only dependency trees. We argue that constituent trees also encode important information, such as explicit span-boundary signals captured by nested multi-word phrases, extra linguistic labels and hierarchical structures useful for detecting anaphora. In this work, we propose a simple yet effective graph-based method to incorporate constituent syntactic structures. Moreover, we also explore to utilise higher-order neighbourhood information to encode rich structures in constituent trees. A novel message propagation mechanism is therefore proposed to enable information flow among elements in syntax trees. Experiments on the English and Chinese portions of OntoNotes 5.0 benchmark show that our proposed model either beats a strong baseline or achieves new state-of-the-art performance. (Code is available at https://github.com/Fantabulous-J/Coref-Constituent-Graph)

Jing Zuo, Lingzhou Mu, Fan Jiang et al.

Achieving human-level performance in Vision-and-Language Navigation (VLN) requires an embodied agent to jointly understand multimodal instructions and visual-spatial context while reasoning over long action sequences. Recent works, such as NavCoT and NavGPT-2, demonstrate the potential of Chain-of-Thought (CoT) reasoning for improving interpretability and long-horizon planning. Moreover, multimodal extensions like OctoNav-R1 and CoT-VLA further validate CoT as a promising pathway toward human-like navigation reasoning. However, existing approaches face critical drawbacks: purely textual CoTs lack spatial grounding and easily overfit to sparse annotated reasoning steps, while multimodal CoTs incur severe token inflation by generating imagined visual observations, making real-time navigation impractical. In this work, we propose FantasyVLN, a unified implicit reasoning framework that preserves the benefits of CoT reasoning without explicit token overhead. Specifically, imagined visual tokens are encoded into a compact latent space using a pretrained Visual AutoRegressor (VAR) during CoT reasoning training, and the model jointly learns from textual, visual, and multimodal CoT modes under a unified multi-CoT strategy. At inference, our model performs direct instruction-to-action mapping while still enjoying reasoning-aware representations. Extensive experiments on LH-VLN show that our approach achieves reasoning-aware yet real-time navigation, improving success rates and efficiency while reducing inference latency by an order of magnitude compared to explicit CoT methods.

Zedong Chu, Shichao Xie, Xiaolong Wu et al.

Embodied navigation has long been fragmented by task-specific architectures. We introduce ABot-N0, a unified Vision-Language-Action (VLA) foundation model that achieves a ``Grand Unification'' across 5 core tasks: Point-Goal, Object-Goal, Instruction-Following, POI-Goal, and Person-Following. ABot-N0 utilizes a hierarchical ``Brain-Action'' architecture, pairing an LLM-based Cognitive Brain for semantic reasoning with a Flow Matching-based Action Expert for precise, continuous trajectory generation. To support large-scale learning, we developed the ABot-N0 Data Engine, curating 16.9M expert trajectories and 5.0M reasoning samples across 7,802 high-fidelity 3D scenes (10.7 $\text{km}^2$). ABot-N0 achieves new SOTA performance across 7 benchmarks, significantly outperforming specialized models. Furthermore, our Agentic Navigation System integrates a planner with hierarchical topological memory, enabling robust, long-horizon missions in dynamic real-world environments.

Mengchao Wang, Qiang Wang, Fan Jiang et al.

Creating a realistic animatable avatar from a single static portrait remains challenging. Existing approaches often struggle to capture subtle facial expressions, the associated global body movements, and the dynamic background. To address these limitations, we propose a novel framework that leverages a pretrained video diffusion transformer model to generate high-fidelity, coherent talking portraits with controllable motion dynamics. At the core of our work is a dual-stage audio-visual alignment strategy. In the first stage, we employ a clip-level training scheme to establish coherent global motion by aligning audio-driven dynamics across the entire scene, including the reference portrait, contextual objects, and background. In the second stage, we refine lip movements at the frame level using a lip-tracing mask, ensuring precise synchronization with audio signals. To preserve identity without compromising motion flexibility, we replace the commonly used reference network with a facial-focused cross-attention module that effectively maintains facial consistency throughout the video. Furthermore, we integrate a motion intensity modulation module that explicitly controls expression and body motion intensity, enabling controllable manipulation of portrait movements beyond mere lip motion. Extensive experimental results show that our proposed approach achieves higher quality with better realism, coherence, motion intensity, and identity preservation. Ours project page: https://fantasy-amap.github.io/fantasy-talking/.

Haowei Zhu, Dehua Tang, Ji Liu et al.

Diffusion models have achieved remarkable progress in the field of image generation due to their outstanding capabilities. However, these models require substantial computing resources because of the multi-step denoising process during inference. While traditional pruning methods have been employed to optimize these models, the retraining process necessitates large-scale training datasets and extensive computational costs to maintain generalization ability, making it neither convenient nor efficient. Recent studies attempt to utilize the similarity of features across adjacent denoising stages to reduce computational costs through simple and static strategies. However, these strategies cannot fully harness the potential of the similar feature patterns across adjacent timesteps. In this work, we propose a novel pruning method that derives an efficient diffusion model via a more intelligent and differentiable pruner. At the core of our approach is casting the model pruning process into a SubNet search process. Specifically, we first introduce a SuperNet based on standard diffusion via adding some backup connections built upon the similar features. We then construct a plugin pruner network and design optimization losses to identify redundant computation. Finally, our method can identify an optimal SubNet through few-step gradient optimization and a simple post-processing procedure. We conduct extensive experiments on various diffusion models including Stable Diffusion series and DiTs. Our DiP-GO approach achieves 4.4 x speedup for SD-1.5 without any loss of accuracy, significantly outperforming the previous state-of-the-art methods.

Wannita Takerngsaksiri, Jirat Pasuksmit, Patanamon Thongtanunam et al.

Recently, Large Language Models (LLMs)-based multi-agent paradigms for software engineering are introduced to automatically resolve software development tasks (e.g., from a given issue to source code). However, existing work is evaluated based on historical benchmark datasets, rarely considers human feedback at each stage of the automated software development process, and has not been deployed in practice. In this paper, we introduce a Human-in-the-loop LLM-based Agents framework (HULA) for software development that allows software engineers to refine and guide LLMs when generating coding plans and source code for a given task. We design, implement, and deploy the HULA framework into Atlassian JIRA for internal uses. Through a multi-stage evaluation of the HULA framework, Atlassian software engineers perceive that HULA can minimize the overall development time and effort, especially in initiating a coding plan and writing code for straightforward tasks. On the other hand, challenges around code quality remain a concern in some cases. We draw lessons learned and discuss opportunities for future work, which will pave the way for the advancement of LLM-based agents in software development.

Yunpeng Zhang, Qiang Wang, Fan Jiang et al.

Tuning-free approaches adapting large-scale pre-trained video diffusion models for identity-preserving text-to-video generation (IPT2V) have gained popularity recently due to their efficacy and scalability. However, significant challenges remain to achieve satisfied facial dynamics while keeping the identity unchanged. In this work, we present a novel tuning-free IPT2V framework by enhancing face knowledge of the pre-trained video model built on diffusion transformers (DiT), dubbed FantasyID. Essentially, 3D facial geometry prior is incorporated to ensure plausible facial structures during video synthesis. To prevent the model from learning copy-paste shortcuts that simply replicate reference face across frames, a multi-view face augmentation strategy is devised to capture diverse 2D facial appearance features, hence increasing the dynamics over the facial expressions and head poses. Additionally, after blending the 2D and 3D features as guidance, instead of naively employing cross-attention to inject guidance cues into DiT layers, a learnable layer-aware adaptive mechanism is employed to selectively inject the fused features into each individual DiT layers, facilitating balanced modeling of identity preservation and motion dynamics. Experimental results validate our model's superiority over the current tuning-free IPT2V methods.

Yu Shen, Shiyang Liu, Qihang He et al.

Big data platforms are widely used in modern enterprises, and an in-production intelligent assistant is increasingly important to help users quickly find actionable guidance and reduce operational burden. While recent LLM+RAG assistants provide a natural interface, they face practical challenges in real deployments: limited scenario coverage across both general consultation and domain-specific troubleshooting workflows, inefficient knowledge access due to inadequate multi-hop retrieval and flat knowledge organization, and high maintenance cost because escalated tickets are unstructured and hard to convert into assistant improvements and reusable SOPs. In this paper, we present SiriusHelper, a deployed intelligent assistant for big data platforms. SiriusHelper serves as a unified online assistant that automatically identifies user intent and routes queries to the right handling path, including dedicated expert workflows for specialized scenarios (e.g., SQL execution diagnosis). To support complex troubleshooting, SiriusHelper combines a DeepSearch-driven mechanism with a priority-based hierarchical knowledge base to enable multi-hop retrieval without context overload, thus improving answer reliability and latency. To reduce expert overhead, SiriusHelper further introduces automated ticket understanding and SOP distillation: it diagnoses the assistant failure reason (e.g., missing knowledge or wrong routing) and extracts domain-specific SOPs to continuously enrich the knowledge base. Experiments and online deployment on Tencent Big Data platform show that SiriusHelper outperforms representative alternatives and reduces online ticket volume by 20.8\%.

Ji Liu, Dehua Tang, Yuanxian Huang et al.

Traditional channel-wise pruning methods by reducing network channels struggle to effectively prune efficient CNN models with depth-wise convolutional layers and certain efficient modules, such as popular inverted residual blocks. Prior depth pruning methods by reducing network depths are not suitable for pruning some efficient models due to the existence of some normalization layers. Moreover, finetuning subnet by directly removing activation layers would corrupt the original model weights, hindering the pruned model from achieving high performance. To address these issues, we propose a novel depth pruning method for efficient models. Our approach proposes a novel block pruning strategy and progressive training method for the subnet. Additionally, we extend our pruning method to vision transformer models. Experimental results demonstrate that our method consistently outperforms existing depth pruning methods across various pruning configurations. We obtained three pruned ConvNeXtV1 models with our method applying on ConvNeXtV1, which surpass most SOTA efficient models with comparable inference performance. Our method also achieves state-of-the-art pruning performance on the vision transformer model.

Fan Jiang, Yu Zhao, Chenyang Lyu et al.

We present Marco-MoE, a suite of fully open multilingual sparse Mixture-of-Experts (MoE) models. Marco-MoE features a highly sparse design in which only around 5\% of the total parameters are activated per input token. This extreme sparsity, combined with upcycling from dense models, enables efficient pre-training on 5T tokens. Our models surpass similarly-sized competitors on English and multilingual benchmarks, achieving a best-in-class performance-to-compute ratio. We further post-train these models to create Marco-MoE-\textsc{Instruct} variants, which surpass the performance of competing models possessing $3$--$14\times$ more activated parameters. Our analysis reveals that Marco-MoE learns structured expert activation patterns shared across related languages, while maintaining highly specialized utilization for linguistically isolated ones. We further show that Marco-MoE allows for scalable language expansion without the interference typical of dense models. To support the community, we disclose our full training datasets, recipes, and model weights.

Peiqin Lin, Chenyang Lyu, Wenjiang Luo et al.

Large language models (LLMs) are now deployed worldwide, inspiring a surge of benchmarks that measure their multilingual and multicultural abilities. However, these benchmarks prioritize generic language understanding or superficial cultural trivia, leaving the evaluation of grounded tasks -- where models must reason within real-world, context-rich scenarios -- largely unaddressed. To fill this gap, we present CulturALL, a comprehensive and challenging benchmark to assess LLMs' multilingual and multicultural competence on grounded tasks. CulturALL is built via a human--AI collaborative framework: expert annotators ensure appropriate difficulty and factual accuracy, while LLMs lighten the manual workload. By incorporating diverse sources, CulturALL ensures comprehensive scenario coverage. Each item is carefully designed to present a high level of difficulty, making CulturALL challenging. CulturALL contains 2,610 samples in 14 languages from 51 regions, distributed across 16 topics to capture the full breadth of grounded tasks. Experiments show that the best LLM achieves 44.48% accuracy on CulturALL, underscoring substantial room for improvement.

Fan Jiang, Tom Drummond, Trevor Cohn

Cross-lingual open domain question answering (CLQA) is a complex problem, comprising cross-lingual retrieval from a multilingual knowledge base, followed by answer generation in the query language. Both steps are usually tackled by separate models, requiring substantial annotated datasets, and typically auxiliary resources, like machine translation systems to bridge between languages. In this paper, we show that CLQA can be addressed using a single encoder-decoder model. To effectively train this model, we propose a self-supervised method based on exploiting the cross-lingual link structure within Wikipedia. We demonstrate how linked Wikipedia pages can be used to synthesise supervisory signals for cross-lingual retrieval, through a form of cloze query, and generate more natural questions to supervise answer generation. Together, we show our approach, \texttt{CLASS}, outperforms comparable methods on both supervised and zero-shot language adaptation settings, including those using machine translation.

Biwen Lei, Yang Li, Xinhai Liu et al.

The creation of high-quality 3D assets, a cornerstone of modern game development, has long been characterized by labor-intensive and specialized workflows. This paper presents Hunyuan3D Studio, an end-to-end AI-powered content creation platform designed to revolutionize the game production pipeline by automating and streamlining the generation of game-ready 3D assets. At its core, Hunyuan3D Studio integrates a suite of advanced neural modules (such as Part-level 3D Generation, Polygon Generation, Semantic UV, etc.) into a cohesive and user-friendly system. This unified framework allows for the rapid transformation of a single concept image or textual description into a fully-realized, production-quality 3D model complete with optimized geometry and high-fidelity PBR textures. We demonstrate that assets generated by Hunyuan3D Studio are not only visually compelling but also adhere to the stringent technical requirements of contemporary game engines, significantly reducing iteration time and lowering the barrier to entry for 3D content creation. By providing a seamless bridge from creative intent to technical asset, Hunyuan3D Studio represents a significant leap forward for AI-assisted workflows in game development and interactive media.

Qiang Wang, Mengchao Wang, Fan Jiang et al.

Producing expressive facial animations from static images is a challenging task. Prior methods relying on explicit geometric priors (e.g., facial landmarks or 3DMM) often suffer from artifacts in cross reenactment and struggle to capture subtle emotions. Furthermore, existing approaches lack support for multi-character animation, as driving features from different individuals frequently interfere with one another, complicating the task. To address these challenges, we propose FantasyPortrait, a diffusion transformer based framework capable of generating high-fidelity and emotion-rich animations for both single- and multi-character scenarios. Our method introduces an expression-augmented learning strategy that utilizes implicit representations to capture identity-agnostic facial dynamics, enhancing the model's ability to render fine-grained emotions. For multi-character control, we design a masked cross-attention mechanism that ensures independent yet coordinated expression generation, effectively preventing feature interference. To advance research in this area, we propose the Multi-Expr dataset and ExprBench, which are specifically designed datasets and benchmarks for training and evaluating multi-character portrait animations. Extensive experiments demonstrate that FantasyPortrait significantly outperforms state-of-the-art methods in both quantitative metrics and qualitative evaluations, excelling particularly in challenging cross reenactment and multi-character contexts. Our project page is https://fantasy-amap.github.io/fantasy-portrait/.

Linhan Fang, Fan Jiang, Ann Mary Toms et al.

Accurate wind speed prediction is crucial for designing and selecting sites for offshore wind farms. This paper investigates the effectiveness of various machine learning models in predicting offshore wind power for a site near the Gulf of Mexico by analyzing meteorological data. After collecting and preprocessing meteorological data, nine different input feature combinations were designed to assess their impact on wind power predictions at multiple heights. The results show that using wind speed as the output feature improves prediction accuracy by approximately 10% compared to using wind power as the output. In addition, the improvement of multi-feature input compared with single-feature input is not obvious mainly due to the poor correlation among key features and limited generalization ability of models. These findings underscore the importance of selecting appropriate output features and highlight considerations for using machine learning in wind power forecasting, offering insights that could guide future wind power prediction models and conversion techniques.

Fanxu Meng, Pingzhi Tang, Fan jiang et al.

Decoder-only models generate tokens autoregressively by caching key/value vectors, but as the cache grows, inference becomes memory-bound. To address this issue, we introduce CLOVER (Cross-Layer Orthogonal Vectors), a novel approach that treats pairs of attention layers as a set of low-rank decompositions. CLOVER applies Singular Value Decomposition (SVD) to the \( Q \)-\( K \) and \( V \)-\( O \) pairs within each attention head. The resulting singular values can either guide pruning or serve as trainable parameters for efficient fine-tuning of all orthogonal vectors. After pruning or fine-tuning, these values are reintegrated into the model without increasing its parameter count. We apply CLOVER to various models, including GPT-2 XL, DeepSeek-V2-Lite, Whisper-Large-v3, Stable Diffusion XL, and LLaMA-3.2-11B-Vision. Our results demonstrate that CLOVER significantly improves pruning efficiency. For instance, the perplexity of pruning 70\% of the \( Q \)-\( K \) pairs in GPT-2 XL is similar to that of pruning just 8\% with vanilla methods. Fine-tuning the singular values further results in a full-rank update, outperforming state-of-the-art methods (LoRA, DoRA, HiRA, and PiSSA) by 7.6\%, 5.5\%, 3.8\%, and 0.7\%, respectively, on eight commonsense tasks for LLaMA-2 7B.

MengChao Wang, Qiang Wang, Fan Jiang et al.

Recent advances in audio-driven portrait animation have demonstrated impressive capabilities. However, existing methods struggle to align with fine-grained human preferences across multiple dimensions, such as motion naturalness, lip-sync accuracy, and visual quality. This is due to the difficulty of optimizing among competing preference objectives, which often conflict with one another, and the scarcity of large-scale, high-quality datasets with multidimensional preference annotations. To address these, we first introduce Talking-Critic, a multimodal reward model that learns human-aligned reward functions to quantify how well generated videos satisfy multidimensional expectations. Leveraging this model, we curate Talking-NSQ, a large-scale multidimensional human preference dataset containing 410K preference pairs. Finally, we propose Timestep-Layer adaptive multi-expert Preference Optimization (TLPO), a novel framework for aligning diffusion-based portrait animation models with fine-grained, multidimensional preferences. TLPO decouples preferences into specialized expert modules, which are then fused across timesteps and network layers, enabling comprehensive, fine-grained enhancement across all dimensions without mutual interference. Experiments demonstrate that Talking-Critic significantly outperforms existing methods in aligning with human preference ratings. Meanwhile, TLPO achieves substantial improvements over baseline models in lip-sync accuracy, motion naturalness, and visual quality, exhibiting superior performance in both qualitative and quantitative evaluations. Ours project page: https://fantasy-amap.github.io/fantasy-talking2/

Fan Jiang, Honglin Yu, Grace Chung et al.

The capabilities of Large Language Models (LLMs) in low-resource languages lag far behind those in English, making their universal accessibility a significant challenge. To alleviate this, we present $\textit{Franken-Adapter}$, a modular language adaptation approach for decoder-only LLMs with embedding surgery. Our method begins by creating customized vocabularies for target languages and performing language adaptation through embedding tuning on multilingual data. These pre-trained embeddings are subsequently integrated with LLMs that have been instruction-tuned on English alignment data to enable zero-shot cross-lingual transfer. Our experiments on $\texttt{Gemma2}$ models with up to 27B parameters demonstrate improvements of up to 20% across 96 languages, spanning both discriminative and generative tasks, with minimal regressions ($<$1%) in English. Further in-depth analysis reveals the critical role of customizing tokenizers in enhancing language adaptation, while boosting inference efficiency. Additionally, we show the versatility of our method by achieving a 14% improvement over a math-optimized LLM across 20 languages, offering a modular solution to transfer reasoning abilities across languages post hoc.

Yixiang Dai, Fan Jiang, Chiyu Wang et al.

High-quality 3D world models are pivotal for embodied intelligence and Artificial General Intelligence (AGI), underpinning applications such as AR/VR content creation and robotic navigation. Despite the established strong imaginative priors, current video foundation models lack explicit 3D grounding capabilities, thus being limited in both spatial consistency and their utility for downstream 3D reasoning tasks. In this work, we present FantasyWorld, a geometry-enhanced framework that augments frozen video foundation models with a trainable geometric branch, enabling joint modeling of video latents and an implicit 3D field in a single forward pass. Our approach introduces cross-branch supervision, where geometry cues guide video generation and video priors regularize 3D prediction, thus yielding consistent and generalizable 3D-aware video representations. Notably, the resulting latents from the geometric branch can potentially serve as versatile representations for downstream 3D tasks such as novel view synthesis and navigation, without requiring per-scene optimization or fine-tuning. Extensive experiments show that FantasyWorld effectively bridges video imagination and 3D perception, outperforming recent geometry-consistent baselines in multi-view coherence and style consistency. Ablation studies further confirm that these gains stem from the unified backbone and cross-branch information exchange.

Jinrui Yang, Fan Jiang, Timothy Baldwin

Language fairness in multilingual information retrieval (MLIR) systems is crucial for ensuring equitable access to information across diverse languages. This paper sheds light on the issue, based on the assumption that queries in different languages, but with identical semantics, should yield equivalent ranking lists when retrieving on the same multilingual documents. We evaluate the degree of fairness using both traditional retrieval methods, and a DPR neural ranker based on mBERT and XLM-R. Additionally, we introduce `LaKDA', a novel loss designed to mitigate language biases in neural MLIR approaches. Our analysis exposes intrinsic language biases in current MLIR technologies, with notable disparities across the retrieval methods, and the effectiveness of LaKDA in enhancing language fairness.

Lingzhou Mu, Qiang Wang, Fan Jiang et al.

Human-Scene Interaction (HSI) seeks to generate realistic human behaviors within complex environments, yet it faces significant challenges in handling long-horizon, high-level tasks and generalizing to unseen scenes. To address these limitations, we introduce FantasyHSI, a novel HSI framework centered on video generation and multi-agent systems that operates without paired data. We model the complex interaction process as a dynamic directed graph, upon which we build a collaborative multi-agent system. This system comprises a scene navigator agent for environmental perception and high-level path planning, and a planning agent that decomposes long-horizon goals into atomic actions. Critically, we introduce a critic agent that establishes a closed-loop feedback mechanism by evaluating the deviation between generated actions and the planned path. This allows for the dynamic correction of trajectory drifts caused by the stochasticity of the generative model, thereby ensuring long-term logical consistency. To enhance the physical realism of the generated motions, we leverage Direct Preference Optimization (DPO) to train the action generator, significantly reducing artifacts such as limb distortion and foot-sliding. Extensive experiments on our custom SceneBench benchmark demonstrate that FantasyHSI significantly outperforms existing methods in terms of generalization, long-horizon task completion, and physical realism. Ours project page: https://fantasy-amap.github.io/fantasy-hsi/

Qionghao Huang, Lingnuo Lu, Xuemei Wu et al.

Adaptive Curriculum Sequencing (ACS) is essential for personalized online learning, yet current approaches struggle to balance complex educational constraints and maintain optimization stability. This paper proposes a Memetic Walrus Optimizer (MWO) that enhances optimization performance through three key innovations: (1) an expert-guided strategy with aging mechanism that improves escape from local optima; (2) an adaptive control signal framework that dynamically balances exploration and exploitation; and (3) a three-tier priority mechanism for generating educationally meaningful sequences. We formulate ACS as a multi-objective optimization problem considering concept coverage, time constraints, and learning style compatibility. Experiments on the OULAD dataset demonstrate MWO's superior performance, achieving 95.3% difficulty progression rate (compared to 87.2% in baseline methods) and significantly better convergence stability (standard deviation of 18.02 versus 28.29-696.97 in competing algorithms). Additional validation on benchmark functions confirms MWO's robust optimization capability across diverse scenarios. The results demonstrate MWO's effectiveness in generating personalized learning sequences while maintaining computational efficiency and solution quality.

Yiding Wang, Fauxu Meng, Xuefeng Zhang et al. · pku

Existing parameter-efficient fine-tuning (PEFT) methods for large language models (LLMs), such as LoRA and PiSSA, constrain model updates to low-rank subspaces, limiting their expressiveness and leading to suboptimal performance on complex tasks. To address this, we introduce High-rank Distributed PiSSA (HD-PiSSA), a distributed PEFT approach that initializes orthogonal adapters across different devices and aggregates their delta updates collectively on W for fine-tuning. Unlike Data Parallel LoRA or PiSSA, which maintain identical adapters across all devices, HD-PiSSA assigns different principal components of the pre-trained weights to each GPU, significantly expanding the range of update directions. This results in over 16x higher effective updated ranks than data-parallel LoRA or PiSSA when fine-tuning on 8 GPUs with the same per-device adapter rank. Empirically, we evaluate HD-PiSSA across various challenging downstream tasks, including mathematics, code generation, and multi-task learning. In the multi-task setting, HD-PiSSA achieves average gains of 10.0 absolute points (14.63%) over LoRA and 4.98 points (6.60%) over PiSSA across 12 benchmarks, demonstrating its benefits from the extra optimization flexibility.

Fan Jiang, Tom Drummond, Trevor Cohn

Recent approaches to multilingual open-domain question answering (MLODQA) have achieved promising results given abundant language-specific training data. However, the considerable annotation cost limits the application of these methods for underrepresented languages. We introduce a \emph{few-shot learning} approach to synthesise large-scale multilingual data from large language models (LLMs). Our method begins with large-scale self-supervised pre-training using WikiData, followed by training on high-quality synthetic multilingual data generated by prompting LLMs with few-shot supervision. The final model, \textsc{FsModQA}, significantly outperforms existing few-shot and supervised baselines in MLODQA and cross-lingual and monolingual retrieval. We further show our method can be extended for effective zero-shot adaptation to new languages through a \emph{cross-lingual prompting} strategy with only English-supervised data, making it a general and applicable solution for MLODQA tasks without costly large-scale annotation.

Kedeng Tong, Xin Jin, Chen Wang et al.

Light field image becomes one of the most promising media types for immersive video applications. In this paper, we propose a novel end-to-end spatial-angular-decorrelated network (SADN) for high-efficiency light field image compression. Different from the existing methods that exploit either spatial or angular consistency in the light field image, SADN decouples the angular and spatial information by dilation convolution and stride convolution in spatial-angular interaction, and performs feature fusion to compress spatial and angular information jointly. To train a stable and robust algorithm, a large-scale dataset consisting of 7549 light field images is proposed and built. The proposed method provides 2.137 times and 2.849 times higher compression efficiency relative to H.266/VVC and H.265/HEVC inter coding, respectively. It also outperforms the end-to-end image compression networks by an average of 79.6% bitrate saving with much higher subjective quality and light field consistency.

Yu Ge, Yibo Wu, Fan Jiang et al.

5G millimeter wave (mmWave) signals have inherent geometric connections to the propagation channel and the propagation environment. Thus, they can be used to jointly localize the receiver and map the propagation environment, which is termed as simultaneous localization and mapping (SLAM). One of the most important tasks in the 5G SLAM is to deal with the nonlinearity of the measurement model. To solve this problem, existing 5G SLAM approaches rely on sigma-point or extended Kalman filters, linearizing the measurement function with respect to the prior probability density function (PDF). In this paper, we study the linearization of the measurement function with respect to the posterior PDF, and implement the iterated posterior linearization filter into the Poisson multi-Bernoulli SLAM filter. Simulation results demonstrate the accuracy and precision improvements of the resulting SLAM filter.

Fan Jiang, Andrew Marmon, Ildebrando De Courten et al.

In many applications of computer vision it is important to accurately estimate the trajectory of an object over time by fusing data from a number of sources, of which 2D and 3D imagery is only one. In this paper, we show how to use a deep feature encoding in conjunction with generative densities over the features in a factor-graph based, probabilistic tracking framework. We present a likelihood model that combines a learned feature encoder with generative densities over them, both trained in a supervised manner. We also experiment with directly inferring probability through the use of image classification models that feed into the likelihood formulation. These models are used to implement deep factors that are added to the factor graph to complement other factors that represent domain-specific knowledge such as motion models and/or other prior information. Factors are then optimized together in a non-linear least-squares tracking framework that takes the form of an Extended Kalman Smoother with a Gaussian prior. A key feature of our likelihood model is that it leverages the Lie group properties of the tracked target's pose to apply the feature encoding on an image patch, extracted through a differentiable warp function inspired by spatial transformer networks. To illustrate the proposed approach we evaluate it on a challenging social insect behavior dataset, and show that using deep features does outperform these earlier linear appearance models used in this setting.

Jiguang He, Fan Jiang, Kamran Keykhosravi et al.

Upcoming beyond fifth generation (5G) communications systems aim at further enhancing key performance indicators and fully supporting brand new use cases by embracing emerging techniques, e.g., reconfigurable intelligent surface (RIS), integrated communication, localization, and sensing, and mmWave/THz communications. The wireless intelligence empowered by state-of-the-art artificial intelligence techniques has been widely considered at the transceivers, and now the paradigm is deemed to be shifted to the smart control of radio propagation environment by virtue of RISs. In this article, we argue that to harness the full potential of RISs, localization and communication must be tightly coupled. This is in sharp contrast to 5G and earlier generations, where localization was a minor additional service. To support this, we first introduce the fundamentals of RIS mmWave channel modeling, followed by RIS channel state information acquisition and link establishment. Then, we deal with the connection between localization and communications, from a separate and joint perspective.

Fan Jiang, Varun Agrawal, Russell Buchanan et al.

State estimation of multi-modal hybrid systems is an important problem with many applications in the field robotics. However, incorporating discrete modes in the estimation process is hampered by a potentially combinatorial growth in computation. In this paper we present a novel incremental multi-hypothesis smoother based on eliminating a hybrid factor graph into a multi-hypothesis Bayes tree, which represents possible discrete state sequence hypotheses. Following iSAM, we enable incremental inference by conditioning the past on the future but we add to that the capability of maintaining multiple discrete mode histories, exploiting the temporal structure of the problem to obtain a simplified representation that unifies the multiple hypothesis tree with the Bayes tree. In the results section we demonstrate the generality of the algorithm with examples in three problem domains: lane change detection (1D), aircraft maneuver detection (2D), and contact detection in legged robots (3D).

Fan Jiang, Zhihui Hao, Xinran Liu

A growing demand for natural-scene text detection has been witnessed by the computer vision community since text information plays a significant role in scene understanding and image indexing. Deep neural networks are being used due to their strong capabilities of pixel-wise classification or word localization, similar to being used in common vision problems. In this paper, we present a novel two-task network with integrating bottom and top cues. The first task aims to predict a pixel-by-pixel labeling and based on which, word proposals are generated with a canonical connected component analysis. The second task aims to output a bundle of character candidates used later to verify the word proposals. The two sub-networks share base convolutional features and moreover, we present a new loss to strengthen the interaction between them. We evaluate the proposed network on public benchmark datasets and show it can detect arbitrary-orientation scene text with a finer output boundary. In ICDAR 2013 text localization task, we achieve the state-of-the-art performance with an F-score of 0.919 and a much better recall of 0.915.

Kai Li, Chau Yuen, Salil S. Kanhere et al.

The main challenges in large-scale people tracking are the recognition of people density in a specific area and tracking the people flow path. To address these challenges, we present SenseFlow, a lightweight people tracking system. SenseFlow utilises off-the-shelf devices which sniff probe requests periodically polled by user's smartphones in a passive manner. We demonstrate the feasibility of SenseFlow by building a proof-of-concept prototype and undertaking extensive evaluations in real-world settings. We deploy the system in one laboratory to study office hours of researchers, a crowded public area in city to evaluate the scalability and performance "in the wild", and four classrooms in the university to monitor the number of students. We also evaluate SenseFlow with varying walking speeds and different models of smartphones to investigate the people flow tracking performance.

Wei Li, Mingquan Qiu, Zhencai Zhu et al.

Rolling element bearing faults in rotating systems are observed as impulses in the vibration signals, which are usually buried in noises. In order to effectively detect the fault of bearings, a novel spectrum searching method is proposed. The structural information of spectrum (SIOS) on a predefined basis is constructed through a searching algorithm, such that the harmonics of impulses generated by faults can be clearly identified and analyzed. Local peaks of the spectrum are located on a certain bin of the basis, and then the SIOS can interpret the spectrum via the number and energy of harmonics related to frequency bins of the basis. Finally bearings can be diagnosed based on the SIOS by identifying its dominant components. Mathematical formulation is developed to guarantee the correct construction of the SISO through searching. The effectiveness of the proposed method is verified with a simulation signal and a benchmark study of bearings.