Min Hu

Lingxiang Xu, Jiaoyun Yang, Min Hu et al.

Long-term memory enables language model agents to support personalized interactions, but it remains unclear when available memories warrant integration into responses. Existing memory evaluations emphasize retrieval accuracy and downstream task utility, while overlooking whether retrieved sensitive memory content is warranted in the current turn. We introduce RBI-Eval, a controlled measurement study built around a probe set that compares model behavior with and without access to sensitive memory under identical benign prompts. We evaluate four base LLMs against a matched no-memory reference across four memory-access settings: full-context exposure and three retrieval systems. Our results reveal substantial behavioral divergence. With memory available, the separation score for sensitive-memory integration decreases by 8.9\%--26.6\% relative to the matched no-memory reference for GPT-5.4-mini, but by 51.1\%--82.9\% for Claude-Sonnet-4.6, DeepSeek-V4-Flash, and Qwen3.5-9B. Control experiments on DeepSeek and GPT-5.4-mini show this effect is specific to sensitive content, rather than general personalization. Retrieval systems reduce exposure but do not eliminate integration once sensitive memory reaches the generator. These findings suggest safe personalization requires memory-aware decisions at both retrieval and generation time.

SII-GAIR, Sand. ai, Ethan Chern et al.

We present daVinci-MagiHuman, an open-source audio-video generative foundation model for human-centric generation. daVinci-MagiHuman jointly generates synchronized video and audio using a single-stream Transformer that processes text, video, and audio within a unified token sequence via self-attention only. This single-stream design avoids the complexity of multi-stream or cross-attention architectures while remaining easy to optimize with standard training and inference infrastructure. The model is particularly strong in human-centric scenarios, producing expressive facial performance, natural speech-expression coordination, realistic body motion, and precise audio-video synchronization. It supports multilingual spoken generation across Chinese (Mandarin and Cantonese), English, Japanese, Korean, German, and French. For efficient inference, we combine the single-stream backbone with model distillation, latent-space super-resolution, and a Turbo VAE decoder, enabling generation of a 5-second 256p video in 2 seconds on a single H100 GPU. In automatic evaluation, daVinci-MagiHuman achieves the highest visual quality and text alignment among leading open models, along with the lowest word error rate (14.60%) for speech intelligibility. In pairwise human evaluation, it achieves win rates of 80.0% against Ovi 1.1 and 60.9% against LTX 2.3 over 2000 comparisons. We open-source the complete model stack, including the base model, the distilled model, the super-resolution model, and the inference codebase.

Sand. ai, Hansi Teng, Hongyu Jia et al.

We present MAGI-1, a world model that generates videos by autoregressively predicting a sequence of video chunks, defined as fixed-length segments of consecutive frames. Trained to denoise per-chunk noise that increases monotonically over time, MAGI-1 enables causal temporal modeling and naturally supports streaming generation. It achieves strong performance on image-to-video (I2V) tasks conditioned on text instructions, providing high temporal consistency and scalability, which are made possible by several algorithmic innovations and a dedicated infrastructure stack. MAGI-1 facilitates controllable generation via chunk-wise prompting and supports real-time, memory-efficient deployment by maintaining constant peak inference cost, regardless of video length. The largest variant of MAGI-1 comprises 24 billion parameters and supports context lengths of up to 4 million tokens, demonstrating the scalability and robustness of our approach. The code and models are available at https://github.com/SandAI-org/MAGI-1 and https://github.com/SandAI-org/MagiAttention. The product can be accessed at https://sand.ai.

Jie Ma, Min Hu, Pinghui Wang et al.

Audio-Visual Question Answering (AVQA) is a complex multi-modal reasoning task, demanding intelligent systems to accurately respond to natural language queries based on audio-video input pairs. Nevertheless, prevalent AVQA approaches are prone to overlearning dataset biases, resulting in poor robustness. Furthermore, current datasets may not provide a precise diagnostic for these methods. To tackle these challenges, firstly, we propose a novel dataset, MUSIC-AVQA-R, crafted in two steps: rephrasing questions within the test split of a public dataset (MUSIC-AVQA) and subsequently introducing distribution shifts to split questions. The former leads to a large, diverse test space, while the latter results in a comprehensive robustness evaluation on rare, frequent, and overall questions. Secondly, we propose a robust architecture that utilizes a multifaceted cycle collaborative debiasing strategy to overcome bias learning. Experimental results show that this architecture achieves state-of-the-art performance on MUSIC-AVQA-R, notably obtaining a significant improvement of 9.32%. Extensive ablation experiments are conducted on the two datasets mentioned to analyze the component effectiveness within the debiasing strategy. Additionally, we highlight the limited robustness of existing multi-modal QA methods through the evaluation on our dataset. We also conduct experiments combining various baselines with our proposed strategy on two datasets to verify its plug-and-play capability. Our dataset and code are available at https://github.com/reml-group/MUSIC-AVQA-R.

Min Hu, Zhizhong Tan, Bin Liu et al.

This study aims to address the challenges of futures price prediction in high-frequency trading (HFT) by proposing a continuous learning factor predictor based on graph neural networks. The model integrates multi-factor pricing theories with real-time market dynamics, effectively bypassing the limitations of existing methods that lack financial theory guidance and ignore various trend signals and their interactions. We propose three heterogeneous tasks, including price moving average regression, price gap regression and change-point detection to trace the short-, intermediate-, and long-term trend factors present in the data. In addition, this study also considers the cross-sectional correlation characteristics of future contracts, where prices of different futures often show strong dynamic correlations. Each variable (future contract) depends not only on its historical values (temporal) but also on the observation of other variables (cross-sectional). To capture these dynamic relationships more accurately, we resort to the spatio-temporal graph neural network (STGNN) to enhance the predictive power of the model. The model employs a continuous learning strategy to simultaneously consider these tasks (factors). Additionally, due to the heterogeneity of the tasks, we propose to calculate parameter importance with mutual information between original observations and the extracted features to mitigate the catastrophic forgetting (CF) problem. Empirical tests on 49 commodity futures in China's futures market demonstrate that the proposed model outperforms other state-of-the-art models in terms of prediction accuracy. Not only does this research promote the integration of financial theory and deep learning, but it also provides a scientific basis for actual trading decisions.

Jie Ma, Pinghui Wang, Zewei Wang et al.

Question answering methods are well-known for leveraging data bias, such as the language prior in visual question answering and the position bias in machine reading comprehension (extractive question answering). Current debiasing methods often come at the cost of significant in-distribution performance to achieve favorable out-of-distribution generalizability, while non-debiasing methods sacrifice a considerable amount of out-of-distribution performance in order to obtain high in-distribution performance. Therefore, it is challenging for them to deal with the complicated changing real-world situations. In this paper, we propose a simple yet effective novel loss function with adaptive loose optimization, which seeks to make the best of both worlds for question answering. Our main technical contribution is to reduce the loss adaptively according to the ratio between the previous and current optimization state on mini-batch training data. This loose optimization can be used to prevent non-debiasing methods from overlearning data bias while enabling debiasing methods to maintain slight bias learning. Experiments on the visual question answering datasets, including VQA v2, VQA-CP v1, VQA-CP v2, GQA-OOD, and the extractive question answering dataset SQuAD demonstrate that our approach enables QA methods to obtain state-of-the-art in- and out-of-distribution performance in most cases. The source code has been released publicly in \url{https://github.com/reml-group/ALO}.

Ziqi Ye, Limin Huang, Yongji Wu et al.

Augmented reality technology has been widely used in industrial design interaction, exhibition guide, information retrieval and other fields. The combination of artificial intelligence and augmented reality technology has also become a future development trend. This project is an AR visualization system for ship detection and recognition based on AI, which mainly includes three parts: artificial intelligence module, Unity development module and Hololens2AR module. This project is based on R3Det algorithm to complete the detection and recognition of ships in remote sensing images. The recognition rate of model detection trained on RTX 2080Ti can reach 96%. Then, the 3D model of the ship is obtained by ship categories and information and generated in the virtual scene. At the same time, voice module and UI interaction module are added. Finally, we completed the deployment of the project on Hololens2 through MRTK. The system realizes the fusion of computer vision and augmented reality technology, which maps the results of object detection to the AR field, and makes a brave step toward the future technological trend and intelligent application.

Qingdi Yu, Zhiwei Cao, Ruihang Wang et al.

Time series forecasting is crucial for applications like resource scheduling and risk management, where multi-step predictions provide a comprehensive view of future trends. Uncertainty Quantification (UQ) is a mainstream approach for addressing forecasting uncertainties, with Conformal Prediction (CP) gaining attention due to its model-agnostic nature and statistical guarantees. However, most variants of CP are designed for single-step predictions and face challenges in multi-step scenarios, such as reliance on real-time data and limited scalability. This highlights the need for CP methods specifically tailored to multi-step forecasting. We propose the Dual-Splitting Conformal Prediction (DSCP) method, a novel CP approach designed to capture inherent dependencies within time-series data for multi-step forecasting. Experimental results on real-world datasets from four different domains demonstrate that the proposed DSCP significantly outperforms existing CP variants in terms of the Winkler Score, achieving a performance improvement of up to 23.59% compared to state-of-the-art methods. Furthermore, we deployed the DSCP approach for renewable energy generation and IT load forecasting in power management of a real-world trajectory-based application, achieving an 11.25% reduction in carbon emissions through predictive optimization of data center operations and controls.

Anbang Wang, Difei Mei, Zhichao Zhang et al.

This paper presents ReverseNER, a method aimed at overcoming the limitation of large language models (LLMs) in zero-shot named entity recognition (NER) tasks, arising from their reliance on pre-provided demonstrations. ReverseNER tackles this challenge by constructing a reliable example library composed of dozens of entity-labeled sentences, generated through the reverse process of NER. Specifically, while conventional NER methods label entities in a sentence, ReverseNER features reversing the process by using an LLM to generate entities from their definitions and subsequently expand them into full sentences. During the entity expansion process, the LLM is guided to generate sentences by replicating the structures of a set of specific \textsl{feature sentences}, extracted from the task sentences by clustering. This expansion process produces dozens of entity-labeled task-relevant sentences. After constructing the example library, the method selects several semantically similar entity-labeled examples for each task sentence as references to facilitate the LLM's entity recognition. We also propose an entity-level self-consistency scoring mechanism to improve NER performance with LLMs. Experiments show that ReverseNER significantly outperforms other zero-shot NER methods with LLMs, marking a notable improvement in NER for domains without labeled data, while declining computational resource consumption.

Xingyou Yin, Ceyao Zhang, Min Hu et al.

Large Language Models (LLMs) have demonstrated effectiveness as zero-shot time series (TS) forecasters. The key challenge lies in tokenizing TS data into textual representations that align with LLMs' pre-trained knowledge. While existing work often relies on fine-tuning specialized modules to bridge this gap, a distinct, yet challenging, paradigm aims to leverage truly off-the-shelf LLMs without any fine-tuning whatsoever, relying solely on strategic tokenization of numerical sequences. The performance of these fully frozen models is acutely sensitive to the textual representation of the input data, as their parameters cannot adapt to distribution shifts. In this paper, we introduce a simple yet highly effective strategy to overcome this brittleness: injecting noise into the raw time series before tokenization. This non-invasive intervention acts as a form of inference-time augmentation, compelling the frozen LLM to extrapolate based on robust underlying temporal patterns rather than superficial numerical artifacts. We theoretically analyze this phenomenon and empirically validate its effectiveness across diverse benchmarks. Notably, to fully eliminate potential biases from data contamination during LLM pre-training, we introduce two novel TS datasets that fall outside all utilized LLMs' pre-training scopes, and consistently observe improved performance. This study provides a further step in directly leveraging off-the-shelf LLMs for time series forecasting.

Min Hu, Yi Wang, Xiaowei Feng et al.

Time-series anomaly detection plays a vital role in monitoring complex operation conditions. However, the detection accuracy of existing approaches is heavily influenced by pattern distribution, existence of multiple normal patterns, dynamical features representation, and parameter settings. For the purpose of improving the robustness and guaranteeing the accuracy, this research combined the strengths of negative selection, unthresholded recurrence plots, and an extreme learning machine autoencoder and then proposed robust anomaly detection for time-series data (RADTD), which can automatically learn dynamical features in time series and recognize anomalies with low label dependency and high robustness. Yahoo benchmark datasets and three tunneling engineering simulation experiments were used to evaluate the performance of RADTD. The experiments showed that in benchmark datasets RADTD possessed higher accuracy and robustness than recurrence qualification analysis and extreme learning machine autoencoder, respectively, and that RADTD accurately detected the occurrence of tunneling settlement accidents, indicating its remarkable performance in accuracy and robustness.

Chuang Wang, Ruimin Hu, Min Hu et al.

Unlike the constraint frontal face condition, faces in the wild have various unconstrained interference factors, such as complex illumination, changing perspective and various occlusions. Facial expressions recognition (FER) in the wild is a challenging task and existing methods can't perform well. However, for occluded faces (containing occlusion caused by other objects and self-occlusion caused by head posture changes), the attention mechanism has the ability to focus on the non-occluded regions automatically. In this paper, we propose a Lossless Attention Model (LLAM) for convolutional neural networks (CNN) to extract attention-aware features from faces. Our module avoids decay information in the process of generating attention maps by using the information of the previous layer and not reducing the dimensionality. Sequentially, we adaptively refine the feature responses by fusing the attention map with the feature map. We participate in the seven basic expression classification sub-challenges of FG-2020 Affective Behavior Analysis in-the-wild Challenge. And we validate our method on the Aff-Wild2 datasets released by the Challenge. The total accuracy (Accuracy) and the unweighted mean (F1) of our method on the validation set are 0.49 and 0.38 respectively, and the final result is 0.42 (0.67 F1-Score + 0.33 Accuracy).

Xiaohua Wang, Muzi Peng, Lijuan Pan et al.

Compared with facial emotion recognition on categorical model, the dimensional emotion recognition can describe numerous emotions of the real world more accurately. Most prior works of dimensional emotion estimation only considered laboratory data and used video, speech or other multi-modal features. The effect of these methods applied on static images in the real world is unknown. In this paper, a two-level attention with two-stage multi-task learning (2Att-2Mt) framework is proposed for facial emotion estimation on only static images. Firstly, the features of corresponding region(position-level features) are extracted and enhanced automatically by first-level attention mechanism. In the following, we utilize Bi-directional Recurrent Neural Network(Bi-RNN) with self-attention(second-level attention) to make full use of the relationship features of different layers(layer-level features) adaptively. Owing to the inherent complexity of dimensional emotion recognition, we propose a two-stage multi-task learning structure to exploited categorical representations to ameliorate the dimensional representations and estimate valence and arousal simultaneously in view of the correlation of the two targets. The quantitative results conducted on AffectNet dataset show significant advancement on Concordance Correlation Coefficient(CCC) and Root Mean Square Error(RMSE), illustrating the superiority of the proposed framework. Besides, extensive comparative experiments have also fully demonstrated the effectiveness of different components.

Kai Hu, Zhijian Ou, Min Hu et al.

Conditional random fields (CRFs) have been shown to be one of the most successful approaches to sequence labeling. Various linear-chain neural CRFs (NCRFs) are developed to implement the non-linear node potentials in CRFs, but still keeping the linear-chain hidden structure. In this paper, we propose NCRF transducers, which consists of two RNNs, one extracting features from observations and the other capturing (theoretically infinite) long-range dependencies between labels. Different sequence labeling methods are evaluated over POS tagging, chunking and NER (English, Dutch). Experiment results show that NCRF transducers achieve consistent improvements over linear-chain NCRFs and RNN transducers across all the four tasks, and can improve state-of-the-art results.