Distilling reasoning paths from teacher to student models via supervised fine-tuning (SFT) provides a shortcut for improving the reasoning ability of the smaller Large Language Models (LLMs). However, the reasoning paths generated by teacher models often reflect only surface-level traces of their underlying authentic reasoning. Insights from cognitive neuroscience suggest that authentic reasoning involves a complex interweaving between meta-reasoning that selects the appropriate sub-problem from multiple candidates, and solving, which addresses the sub-problem. It means that authentic reasoning has implicit multi-branch structure. Supervised fine-tuning collapses this rich structure into a flat sequence of token prediction in teacher's reasoning path, which cannot distill this structure to student. To address this limitation, we propose RLKD, a reinforcement learning (RL)-based distillation framework guided by a novel Generative Structure Reward Model (GSRM). Our GSRM converts the reasoning path into multiple meta-reasoning-solving steps and gives the reward to measure the alignment between the reasoning structures of student and teacher. Our RLKD combines this reward with RL, enables the student LLM to internalize the teacher’s implicit multi-branch structure in authentic reasoning, rather than merely mimicking fixed teacher's output paths. Experiments show that RLKD, even when trained on only 0.1% of the data under an RL-only regime, surpasses the performance of standard SFT-RL pipelines and further unleashes the potential reasoning ability of the student LLM than SFT-based distillation.

How far are deep models from real-world video anomaly understanding (VAU)? Current works typically emphasize detecting unexpected occurrences deviating from normal patterns or comprehending anomalous events with interpretable descriptions. However, they exhibit only a superficial comprehension of real-world anomalies, with limited breadth in complex principles and subtle contexts that distinguish the anomalies from normalities, e.g., climbing cliffs with safety gear vs. without it. To this end, we introduce CueBench, the first of its kind Benchmark, devoted to Context-aware video anomalies within a Unified Evaluation framework. We comprehensively establish an event-centric hierarchical taxonomy that anchors two core event types: 14 conditional and 18 absolute anomaly events, defined by their refined semantics from diverse contexts across 174 scenes and 198 attributes. Based on this, we propose to unify and benchmark context-aware VAU with various challenging tasks across recognition, temporal grounding, detection, and anticipation. It also serves as a rigorous and fair probing evaluation suite for generalized and specialized vision-language models (VLMs) across both generative and discriminative paradigms. To address the challenges underlying CueBench, we further develop Cue-R1 based on R1-style reinforcement fine-tuning with verifiable, task-aligned, and hierarchy-refined rewards in a unified generative manner. Extensive results on CueBench reveal that, existing VLMs are still far from satisfactory real-world anomaly understanding, while our Cue-R1 surpasses these state-of-the-art approaches by over 24% on average.

Zihao Wei, Liang Pang, Jiahao Liu, Wenjie Shi, Jingcheng Deng, Shicheng Xu, Zenghao Duan, Jingang Wang, Fei Sun, Huawei Shen, Xueqi Cheng. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Implementing federated learning (FL) over wireless networks faces critical research challenges, such as high communication costs, inevitable communication latency, and significant energy consumption for model transmission and training, primarily caused by device heterogeneity and unpredictable dynamic channel conditions. This paper proposes Graph-based Resource Optimization with Compression for FL (GROC-FL), a unified framework that jointly coordinates wireless resource allocation and collaborative model compression. By leveraging Graph Neural Networks (GNNs) to model wireless topology and Deep Reinforcement Learning (DRL) to optimize communication and computation resources together with a globally consistent sparse update mechanism, GROC-FL minimizes the overall energy consumption of clients over wireless networks. In order to address the intrinsic topology dependence of wireless FL, we develop a graph-augmented DRL agent based on a graph convolutional network (GCN) that captures resource competition and network topology. We further develop a collaborative model compression module, termed Federated Parameter Negotiation (FPN), which enables clients to negotiate a global sparse mask and further reduce energy consumption during FL training. Experimental results demonstrate that GROC-FL outperforms the baselines in energy consumption, training performance, and client fairness.

Hanxing Ding, Shuchang Tao, Liang Pang, Zihao Wei, Jinyang Gao, Bolin Ding, Huawei Shen, Xueqi Cheng. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The extensive utilization of large language models (LLMs) underscores the crucial necessity for precise and contemporary knowledge embedded within their intrinsic parameters. Existing research on knowledge editing primarily concentrates on monolingual scenarios, neglecting the complexities presented by multilingual contexts and multi-hop reasoning. To address these challenges, our study introduces MLaKE (Multilingual Language Knowledge Editing), a novel benchmark comprising 4072 multi-hop and 5360 single-hop questions designed to evaluate the adaptability of knowledge editing methods across five languages: English, Chinese, Japanese, French, and German. MLaKE aggregates fact chains from Wikipedia across languages and utilizes LLMs to generate questions and answer. We assessed the effectiveness of current multilingual knowledge editing methods using the MLaKE dataset. Our results show that due to considerable inconsistencies in both multilingual performance and encoding efficiency, these methods struggle to generalize effectively across languages. The accuracy of these methods when editing English is notably higher than for other languages. The experimental results further demonstrate that models encode knowledge and generation capabilities for different languages using distinct parameters, leading to poor cross-lingual transfer performance in current methods. Transfer performance is notably better within the same language family compared to across different families. These findings emphasize the urgent need to improve multilingual knowledge editing methods.

Jingcheng Deng, Zhongtao Jiang, Liang Pang, Zihao Wei, Liwei Chen, Kun Xu, Yang Song, Huawei Shen, Xueqi Cheng. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Current super-resolution (SR) algorithms rely heavily on annotated data and often ignore the uncertainty in image degradation and features, limiting their real-world application. We propose an uncertainty-aware SR network using a self-convolutional attention mechanism. Our approach focuses on an SR reconstruction network enhanced by a cross-scale self-convolutional attention mechanism within the Transformer framework, which leverages local regions at various resolutions as convolution kernels to enhance high-frequency details. We also design a heteroscedastic uncertainty loss function to learn pixel and feature uncertainties, guiding the network to improve textures and edges adaptively. Extensive experiments show that our method achieves superior visual reconstruction on standard real-world datasets.

Recent knowledge editing methods have primarily focused on modifying structured knowledge in large language models. However, this task setting overlooks the fact that a significant portion of real-world knowledge is stored in an unstructured format, characterized by long-form content, noise, and a complex yet comprehensive nature. Techniques like "local layer key-value storage" and "term-driven optimization", as used in previous methods like MEMIT, are not effective for handling unstructured knowledge. To address these challenges, we propose a novel Unstructured Knowledge Editing method, namely UnKE, which extends previous assumptions in the layer dimension and token dimension. Firstly, in the layer dimension, we propose non-local block key-value storage to replace local layer key-value storage, increasing the representation ability of key-value pairs and incorporating attention layer knowledge. Secondly, in the token dimension, we replace "term-driven optimization" with "cause-driven optimization", which edits the last token directly while preserving context, avoiding the need to locate terms and preventing the loss of context information. Results on newly proposed unstructured knowledge editing dataset (UnKEBench) and traditional structured datasets demonstrate that UnKE achieves remarkable performance, surpassing strong baselines. In addition, UnKE has robust batch editing and sequential editing capabilities.

Volumetric optical microscopy using non-diffracting beams enables rapid imaging of 3D volumes by projecting them axially to 2D images but lacks crucial depth information. Addressing this, we introduce MicroDiffusion, a pi-oneering tool facilitating high-quality, depth-resolved 3D volume reconstruction from limited 2D projections. While existing Implicit Neural Representation (INR) models often yield incomplete outputs and Denoising Diffusion Prob-abilistic Models (DDPM) excel at capturing details, our method integrates INR's structural coherence with DDPM's fine-detail enhancement capabilities. We pretrain an INR model to transform 2D axially-projected images into a pre-liminary 3D volume. This pretrained INR acts as a global prior guiding DDPM's generative process through a linear interpolation between INR outputs and noise inputs. This strategy enriches the diffusion process with structured 3D information, enhancing detail and reducing noise in localized 2D images. By conditioning the diffusion model on the closest 2D projection, MicroDiffusion substantially enhances fidelity in resulting 3D reconstructions, surpassing INR and standard DDPM outputs with unparalleled image quality and structural fidelity. Our code and dataset are available at https://github.com/UCSC-VLAA/MicroDiffusion.

We propose a simple strategy for masking image patches during visual-language contrastive learning that improves the quality of the learned representations and the training speed. During each iteration of training, we randomly mask clusters of visually similar image patches, as measured by their raw pixel intensities. This provides an extra learning signal, beyond the contrastive training itself, since it forces a model to predict words for masked visual structures solely from context. It also speeds up training by reducing the amount of data used in each image. We evaluate the effectiveness of our model by pre-training on a number of bench-marks, finding that it outperforms other masking strategies, such as FLIP, on the quality of the learned representation.

Buffer-aided relay technique can improve the diversity order and offer secrecy provision. To further improve secrecy performance, this paper proposes a secure link selection for relay networks where a new link selection policy is first designed under the constraint on the buffers and channel states using a Markov chain. The stationary state and the corresponding state transition matrix can be derived, and they are used to analyze the secrecy performance. Through the derivation of secrecy outage probability, we can get its closed-form expressions. Numerical results demonstrate that the proposed secure transmission scheme has a better performance than the conventional buffer-aided secure transmission schemes in terms of secrecy outage probability.

Since the COVID-19 epidemic is still expanding around the world and poses a serious threat to human life and health, it is necessary for us to carry out epidemic transmission prediction, whole genome sequence analysis, and public psychological stress assessment for 2019-nCoV. However, transmission prediction models are insufficiently accurate and genome sequence characteristics are not clear, and it is difficult to dynamically assess the public psychological stress state under the 2019-nCoV epidemic. Therefore, this study develops a 2019nCoVAS web service (http://www.combio-lezhang.online/2019ncov/home.html) that not only offers online epidemic transmission prediction and lineage-associated underrepresented permutation (LAUP) analysis services to investigate the spreading trends and genome sequence characteristics, but also provides psychological stress assessments based on such an emotional dictionary that we built for 2019-nCoV. Finally, we discuss the shortcomings and further study of the 2019nCoVAS web service.

GIMLI is a 384-bit permutation proposed by Bernstein et al. at CHES 2017. It is designed with the goal of achieving both high security and high performance across a wide range of hardware and software platforms. Since GIMLI can be used as a building block for many cryptographic schemes, it is important to understand its concrete security. To the best of our knowledge, third party cryptanalysis of GIMLI is limited. In this paper, we identify some zero-sum distinguishers for 14-round GIMLI with the inside-out technique, which are one-round longer than the integral distinguishers presented by the designers. Although we obtain improved cryptanalysis results, these zero-sum distinguishers are far from threatening the full version of GIMLI.

Bursty traffic and thousands of concurrent flows incur inevitable congestion in data center networks (DCNs) and then affect the overall performance. Various transport protocols are developed to mitigate the network congestion, including reactive and proactive protocols. Reactive schemes to handling congestion after congestion arises are common to current DCNs. However, with the growth of scale and link speed, reactive schemes such as DCTCP encounter the significant problem of slow responding to congestion. On the contrary, proactive protocols are designed to avoid congestion, and they have the advantages of zero data loss, fast convergence and low buffer occupancy (e.g., credit-reservation protocols). But in actual deployment scenario, it is hard to guarantee one protocol to be deployed in every server at one time. When credit-reservation protocol is deployed to DCNs step-by-step, the network is converted to multi-protocol state and faces the following fundamental challenges: (i) unfairness, (ii) high bu er occupancy, and (iii) heavy tail delay. Therefore, we propose EC4, which is for converging ECN-based and credit-reservation protocols with minimal modification. To the best of our knowledge, EC4is the first to address how to harmonize proactive and reactive congestion control. Targeting the common ECN-based protocol-DCTCP, EC4leverages the Forward Explicit Congestion Notification (FECN) to deliver realtime congestion information and redefines feedback control. After evaluation, the results show that EC4e ectively addresses the unfair link allocation. Furthermore, even workloads at 0.6 does not cause buffer overflow, thus largely eliminating the timeouts problem.