Time series analysis faces significant challenges in handling variable-length data and achieving robust generalization. While Transformer-based models have advanced time series tasks, they often struggle with feature redundancy and limited generalization capabilities. Drawing inspiration from classical CNN architectures' pyramidal structure, we propose a Multi-Scale Representation Learning Framework based on a Conv-like ScaleFusion Transformer. Our approach introduces a temporal convolution-like structure that combines patching operations with multi-head attention, enabling progressive temporal dimension compression and feature channel expansion. We further develop a novel cross-scale attention mechanism for effective feature fusion across different temporal scales, along with a log-space normalization method for variable-length sequences. Extensive experiments demonstrate that our framework achieves superior feature independence, reduced redundancy, and better performance in forecasting and classification tasks compared to state-of-the-art methods.

Diabetes mellitus is a growing global health issue, with Type 1 Diabetes (T1D) requiring constant monitoring to avoid hypoglycemia. Although Continuous Glucose Monitors (CGMs) are effective, their cost and invasiveness limit access, particularly in low-resource settings. This paper proposes a non-invasive method to classify glycemic states using Galvanic Skin Response (GSR), a biosignal commonly captured by wearable sensors. We use the merged OhioT1DM 2018 and 2020 datasets to build a machine learning pipeline that detects hypoglycemia (glucose < 70 mg/dl) and normoglycemia (glucose > 70 mg/dl) with GSR alone. Seven models are trained and evaluated: Random Forest, XGBoost, MLP, CNN, LSTM, Logistic Regression, and K-Nearest Neighbors. Validation sets and 95% confidence intervals are reported to increase reliability and assess robustness. Results show that the LSTM model achieves a perfect hypoglycemia recall (1.00) with an F1-score confidence interval of [0.611-0.745], while XGBoost offers strong performance with a recall of 0.54 even under class imbalance. This approach highlights the potential for affordable, wearable-compatible glucose monitoring tools suitable for settings with limited CGM availability using GSR data. Index Terms: Hypoglycemia Detection, Galvanic Skin Response, Non Invasive Monitoring, Wearables, Machine Learning, Confidence Intervals.

Accurate prediction of road accidents remains challenging due to intertwined spatial, temporal, and contextual factors in urban traffic. We propose MSGAT-GRU, a multi-scale graph attention and recurrent model that jointly captures localized and long-range spatial dependencies while modeling sequential dynamics. Heterogeneous inputs, such as traffic flow, road attributes, weather, and points of interest, are systematically fused to enhance robustness and interpretability. On the Hybrid Beijing Accidents dataset, MSGAT-GRU achieves an RMSE of 0.334 and an F1-score of 0.878, consistently outperforming strong baselines. Cross-dataset evaluation on METR-LA under a 1-hour horizon further supports transferability, with RMSE of 6.48 (vs. 7.21 for the GMAN model) and comparable MAPE. Ablations indicate that three-hop spatial aggregation and a two-layer GRU offer the best accuracy-stability trade-off. These results position MSGAT-GRU as a scalable and generalizable model for intelligent transportation systems, providing interpretable signals that can inform proactive traffic management and road safety analytics.

Irregular multivariate time series (IMTS) is characterized by the lack of synchronized observations across its different channels. In this paper, we point out that this channel-wise asynchrony can lead to poor channel-wise modeling of existing deep learning methods. To overcome this limitation, we propose MTM, a multi-scale token mixing transformer for the classification of IMTS. We find that the channel-wise asynchrony can be alleviated by down-sampling the time series to coarser timescales, and propose to incorporate a masked concat pooling in MTM that gradually down-samples IMTS to enhance the channel-wise attention modules. Meanwhile, we propose a novel channel-wise token mixing mechanism which proactively chooses important tokens from one channel and mixes them with other channels, to further boost the channel-wise learning of our model. Through extensive experiments on real-world datasets and comparison with state-of-the-art methods, we demonstrate that MTM consistently achieves the best performance on all the benchmarks, with improvements of up to 3.8% in AUPRC for classification.

Medical time series (MedTS) classification is pivotal for intelligent healthcare, yet its efficacy is severely limited by poor cross-subject generation due to the profound cross-individual heterogeneity. Despite advances in architectural innovations and transfer learning techniques, current methods remain constrained by modality-specific inductive biases that limit their ability to learn universally invariant representations. To overcome this, we propose TS-P$^2$CL, a novel plug-and-play framework that leverages the universal pattern recognition capabilities of pre-trained vision models. We introduce a vision-guided paradigm that transforms 1D physiological signals into 2D pseudo-images, establishing a bridge to the visual domain. This transformation enables implicit access to rich semantic priors learned from natural images. Within this unified space, we employ a dual-contrastive learning strategy: intra-modal consistency enforces temporal coherence, while cross-modal alignment aligns time-series dynamics with visual semantics, thereby mitigating individual-specific biases and learning robust, domain-invariant features. Extensive experiments on six MedTS datasets demonstrate that TS-P$^2$CL consistently outperforms fourteen methods in both subject-dependent and subject-independent settings.

Conversational agents deployed in industrial-scale official account platforms must generate responses that are both contextually grounded and stylistically aligned-requirements that existing methods struggle to meet. Chain-of-thought (CoT) prompting induces significant latency due to multi-turn reasoning; per-account fine-tuning is computationally prohibitive; and long prompt-based methods degrade the model's ability to grasp injected context and style. In this paper, we propose WeStar, a lite-adaptive framework for stylized contextual question answering that scales to millions of official accounts. WeStar combines context-grounded generation via RAG with style-aware generation using Parametric RAG (PRAG), where LoRA modules are dynamically activated per style cluster. Our contributions are fourfold: (1) We introduce WeStar, a unified framework capable of serving large volumes of official accounts with minimal overhead. (2) We propose a multi-dimensional, cluster-based parameter sharing scheme that enables compact style representation while preserving stylistic diversity. (3) We develop a style-enhanced Direct Preference Optimization (SeDPO) method to optimize each style cluster's parameters for improved generation quality. (4) Experiments on a large-scale industrial dataset validate the effectiveness and efficiency of WeStar, underscoring its pracitical value in real-world deployment.

Large Language Models (LLMs) struggle with information forgetting and inefficiency in long-horizon, multi-turn dialogues. To address this, we propose a training-free prompt engineering method, the State-Update Multi-turn Dialogue Strategy. It utilizes "State Reconstruction" and "History Remind" mechanisms to effectively manage dialogue history. Our strategy shows strong performance across multiple multi-hop QA datasets. For instance, on the HotpotQA dataset, it improves the core information filtering score by 32.6%, leading to a 14.1% increase in the downstream QA score, while also reducing inference time by 73.1% and token consumption by 59.4%. Ablation studies confirm the pivotal roles of both components. Our work offers an effective solution for optimizing LLMs in long-range interactions, providing new insights for developing more robust Agents.

We present Qwen3-Omni, a single multimodal model that, for the first time, maintains state-of-the-art performance across text, image, audio, and video without any degradation relative to single-modal counterparts. Qwen3-Omni matches the performance of same-sized single-modal models within the Qwen series and excels particularly on audio tasks. Across 36 audio and audio-visual benchmarks, Qwen3-Omni achieves open-source SOTA on 32 benchmarks and overall SOTA on 22, outperforming strong closed-source models such as Gemini-2.5-Pro, Seed-ASR, and GPT-4o-Transcribe. Qwen3-Omni adopts a Thinker-Talker MoE architecture that unifies perception and generation across text, images, audio, and video, yielding fluent text and natural real-time speech. It supports text interaction in 119 languages, speech understanding in 19 languages, and speech generation in 10 languages. To reduce first-packet latency in streaming synthesis, Talker autoregressively predicts discrete speech codecs using a multi-codebook scheme. Leveraging the representational capacity of these codebooks, we replace computationally intensive block-wise diffusion with a lightweight causal ConvNet, enabling streaming from the first codec frame. In cold-start settings, Qwen3-Omni achieves a theoretical end-to-end first-packet latency of 234 ms. To further strengthen multimodal reasoning, we introduce a Thinking model that explicitly reasons over inputs from any modality. Since the research community currently lacks a general-purpose audio captioning model, we fine-tuned Qwen3-Omni-30B-A3B to obtain Qwen3-Omni-30B-A3B-Captioner, which produces detailed, low-hallucination captions for arbitrary audio inputs. Qwen3-Omni-30B-A3B, Qwen3-Omni-30B-A3B-Thinking, and Qwen3-Omni-30B-A3B-Captioner are publicly released under the Apache 2.0 license.

In particular, we contrasted direct supervision based on antideriva-tive estimates with several method classes that rely on repeated differentiation, including both automatic and numerical approaches. As part of this framework, we explored integral reduction techniques as a means to mitigate the computational overhead associated with repeated integration or differentiation. Our systematic experimental analysis of all presented methods across a diverse set of modalities reveals that differential supervision via naïve automatic differentiation generally outperforms all competing approaches in terms of result quality. However, this performance comes at the cost of substantial training time, particularly for higher-order integration. A practical alternative in many scenarios is numerical differentiation via finite differences, which often achieves a reasonable trade-off between quality and computational cost. Nevertheless, it is important to incorporate a compensation operator to correct for the signal smoothing inherent in this differentiation scheme. Our findings suggest directions for future research, such as investigating progressive supervision paradigms that leverage different supervisory signals throughout the learning process. Interestingly, our analysis reveals that antiderivative quality correlates only weakly with downstream task performance. Cumulative schemes often excel at efficiently computing highly non-local aggregates, where local inaccuracies tend to cancel out.

This is especially challenging in Class-Imbalanced Multi-Label Image Classification (CI-MLIC) tasks, where data imbalance and multi-object recognition present significant obstacles. T o address these challenges, we propose a novel method termed Dual-View Alignment Learning with Hierarchical Prompt (HP-DV AL), which leverages multi-modal knowledge from vision-language pretrained (VLP) models to mitigate the class-imbalance problem in multi-label settings. Specifically, HP-DV AL employs dual-view alignment learning to transfer the powerful feature representation capabilities from VLP models by extracting complementary features for accurate image-text alignment. T o better adapt VLP models for CI-MLIC tasks, we introduce a hierarchical prompt-tuning strategy that utilizes global and local prompts to learn task-specific and context-related prior knowledge. Additionally, we design a semantic consistency loss during prompt tuning to prevent learned prompts from deviating from general knowledge embedded in VLP models. The effectiveness of our approach is validated on two CI-MLIC benchmarks: MS-COCO and VOC2007. Extensive experimental results demonstrate the superiority of our method over SOT A approaches, achieving mAP improvements of 10.0% and 5.2% on the long-tailed multi-label image classification task, and 6.8% and 2.9% on the multi-label few-shot image classification task.