arXiv subjects

Kai Xie

Publications and source records attributed to Kai Xie.

At least 19 recordsLinked to original sources

arXiv2026

Echo-Aware Modulation for Compact-Latent Frequency-Time Modeling in Lightweight Acoustic Echo Cancellation

Existing lightweight acoustic echo cancellation (AEC) systems often combine linear AEC with Bark-domain DNN-based suppression to lower the computational footprint. In such systems, downsampling layers further compress the input features into a compact bottleneck representation, but this compression weakens frequency-time modeling capacity and degrades performance. To mitigate this limitation, we propose MSA-EchoLite, a lightweight Bark-domain AEC framework with an asymmetric dual-branch encoder and an echo-aware frequency-time modulation (EAM) module. The EAM module enriches the compressed bottleneck representation by modeling discrepancy and correlation cues between the dual-branch microphone and echo-related latent features. Experimental results show that the Bark-domain variant of MSA-EchoLite offers a better performance-complexity trade-off than its frequency-domain counterpart but is more sensitive to feature compression. With only 26.1% additional FLOPs over its non-EAM Bark-domain variant, its EAM-enhanced version achieves 99.1% of the PESQ of the frequency-domain counterpart, which requires nearly twice the FLOPs, and even surpasses it in SDR. Overall, MSA-EchoLite outperforms state-of-the-art lightweight AEC models while using only 0.2 M parameters and 100 M FLOPs/s.

eess.AS↗

arXiv2026

Budget-Aware LLM Discovery via Cost-Calibrated Frontier Utility

Large language models increasingly support scientific and algorithmic discovery through inference-time search over evaluated candidates. Existing adaptive discovery controllers assign credit based only on score progress, even though prompt length, retries, and guidance calls cause search actions to incur different token costs. We prove that cost-blind credit can forfeit all but a vanishing fraction of attainable quality as frontiers multiply and costs diverge. Under a fixed search-side token budget, the controller must decide which frontier is improving and whether its gain justifies the realized cost before the budget is exhausted. We introduce \textbf{CostAda}, a cost-calibrated adaptive controller built around \emph{cost-calibrated frontier utility}. The utility values frontier progress relative to realized action cost and conditions that credit on the remaining budget. CostAda uses this signal to control local exploration intensity, frontier allocation, and budgeted tactic intervention. Cost and remaining budget therefore shape the search rather than serving only as accounting variables or a stopping rule. CostAda reaches the strongest baseline's full-budget quality with at most half the budget on twelve of sixteen benchmark--backbone pairs while achieving the strongest mean final quality on all eight benchmarks under GLM-5 and GPT-5.4.

cs.LG↗

arXiv2026

ICME 2026 Grand Challenge on Cross-Scenario Defect Detection and Fine-Grained Severity Grading for High-Precision Manufacturing

This paper presents the IEEE International Conference on Multimedia and Expo (ICME) 2026 Grand Challenge on Cross-Scenario Defect Detection and Fine-Grained Severity Grading for High-Precision Manufacturing. The challenge is motivated by two key limitations of existing industrial defect inspection systems: (1) current deep learning-based methods often suffer significant performance degradation when deployed in unseen production scenarios, and (2) most benchmarks neglect severity-aware assessment, which is critical for risk control and yield optimization. To address these limitations, we design two complementary tracks: Track 1 (Cross-Scenario Defect Detection) targets accurate defect detection, localization, and classification across diverse unseen production environments; Track 2 (Fine-Grained Severity Grading) requires assigning each detected defect an industry-standard severity level, including Acceptable, Marginal NG, NG, and Gross NG. We construct a large-scale industrial dataset of high-resolution microscopic images spanning seven representative defect categories, comprising over 3,800 images with pixel-level instance annotations for Track 1 and over 2,600 images with severity-grade labels for Track 2. The challenge attracted 86 registered participants with 130 submissions; during the final testing phase, 21 teams submitted results and 12 teams provided models with technical reports. The resulting benchmark, together with the diverse and effective solutions contributed by participating teams, sets a new standard for industrial defect analysis research.

cs.CV↗

arXiv2026

DSWAM: A Dual-System World Action Foundation Model for Fine-Grained Robot Manipulation

World Action Models (WAMs) provide a promising alternative to Vision-Language-Action (VLA) policies by using video-based world modeling as dense supervision for robot action learning. Existing WAMs excel at physically grounded execution, but typically lack the explicit language-level planning interface in VLM-based VLAs for decomposing coarse instructions. Such decomposition becomes important when household tasks involve complex multi-step goals, where coarse user commands need to be converted into sequences of fine-grained executable subtasks. Meanwhile, the field still lacks a fair real-robot comparison between VLA and WAM execution capabilities, since existing systems often differ in data, robot embodiments, and task protocols. To address both the decomposition gap and the need for a controlled WAM-VLA comparison, we introduce DSWAM, a Dual-System World Action Foundation Model for fine-grained robot manipulation. DSWAM keeps a System 1 WAM executor as the default control path and optionally activates a System 2 vision-language subtask planner only when task decomposition is useful. The planner predicts executable subtasks from short-term visual history and a global task prompt, while the WAM executor performs world-aware action generation for each instruction or subtask. The executor is trained with action prediction and video co-training, but inference directly predicts action chunks without explicit future video generation. To make this execution path practical on real robots, we further integrate TensorRT acceleration, asynchronous execution, and real-time chunking (RTC) so that policy queries do not block robot control. To provide a fair real-robot comparison with VLA policies, we build and evaluate DSWAM under the DeMaVLA real-world deformable manipulation setting with matched robot platform, pretraining data, post-training data, and evaluation criteria.

cs.RO↗

arXiv2026

Deep CSI Feedback for FDD Massive MIMO Systems: A Curvelet Learning Approach

Downlink channel state information (CSI) feedback plays a key role in frequency division duplex (FDD) massive multiple-input multiple-output (mMIMO) systems. The growth of antennas in ultra-massive MIMO increases the difficulty and overhead of CSI feedback, which poses significant challenges for conventional downlink CSI feedback mechanisms. To address the limitations of existing CSI feedback approaches, this paper proposes a novel curvelet learning based framework termed SwinCANet, comprising a frequency-domain information processing module and a denoising module. The frequency-domain information processing module employs curvelet transform to decompose CSI into low-frequency and high-frequency components. Subsequently, Swin Transformer and channel-wise attention block are utilized for extracting the low-frequency and high-frequency representations, respectively, thereby enhancing reconstruction quality. Notably, an additional Swin Transformer facilitates the fusion of multi-scale frequency components, enhancing capabilities across different angular resolutions and spatial directions. Furthermore, we develop a variant (De-SwinCANet), which employs a Sigmoid threshold function to effectively suppress noise coefficients, thereby mitigating various channel impairments and nonlinear distortions. Numerical simulation results demonstrate that the proposed methodology achieves superior performance compared to existing benchmarks while maintaining robust performance under challenging propagation conditions.

eess.SP↗

arXiv2026

Mozi: Governed Autonomy for Drug Discovery LLM Agents

Tool-augmented large language model (LLM) agents promise to unify scientific reasoning with computation, yet their deployment in high-stakes domains like drug discovery is bottlenecked by two critical barriers: unconstrained tool-use governance and poor long-horizon reliability. In dependency-heavy pharmaceutical pipelines, autonomous agents often drift into irreproducible trajectories, where early-stage hallucinations multiplicatively compound into downstream failures. To overcome this, we present Mozi, a dual-layer architecture that bridges the flexibility of generative AI with the deterministic rigor of computational biology. Layer A (Control Plane) establishes a governed supervisor--worker hierarchy that enforces role-based tool isolation, limits execution to constrained action spaces, and drives reflection-based replanning. Layer B (Workflow Plane) operationalizes canonical drug discovery stages -- from Target Identification to Lead Optimization -- as stateful, composable skill graphs. This layer integrates strict data contracts and strategic human-in-the-loop (HITL) checkpoints to safeguard scientific validity at high-uncertainty decision boundaries. Operating on the design principle of ``free-form reasoning for safe tasks, structured execution for long-horizon pipelines,'' Mozi provides built-in robustness mechanisms and trace-level audibility to completely mitigate error accumulation. We evaluate Mozi on PharmaBench, a curated benchmark for biomedical agents, demonstrating superior orchestration accuracy over existing baselines. Furthermore, through end-to-end therapeutic case studies, we demonstrate Mozi's ability to navigate massive chemical spaces, enforce stringent toxicity filters, and generate highly competitive in silico candidates, effectively transforming the LLM from a fragile conversationalist into a reliable, governed co-scientist.

cs.AI↗

arXiv2026

LEMAS: Large A 150K-Hour Large-scale Extensible Multilingual Audio Suite with Generative Speech Models

We present the LEMAS-Dataset, which, to our knowledge, is currently the largest open-source multilingual speech corpus with word-level timestamps. Covering over 150,000 hours across 10 major languages, LEMAS-Dataset is constructed via a efficient data processing pipeline that ensures high-quality data and annotations. To validate the effectiveness of LEMAS-Dataset across diverse generative paradigms, we train two benchmark models with distinct architectures and task specializations on this dataset. LEMAS-TTS, built upon a non-autoregressive flow-matching framework, leverages the dataset's massive scale and linguistic diversity to achieve robust zero-shot multilingual synthesis. Our proposed accent-adversarial training and CTC loss mitigate cross-lingual accent issues, enhancing synthesis stability. Complementarily, LEMAS-Edit employs an autoregressive decoder-only architecture that formulates speech editing as a masked token infilling task. By exploiting precise word-level alignments to construct training masks and adopting adaptive decoding strategies, it achieves seamless, smooth-boundary speech editing with natural transitions. Experimental results demonstrate that models trained on LEMAS-Dataset deliver high-quality synthesis and editing performance, confirming the dataset's quality. We envision that this richly timestamp-annotated, fine-grained multilingual corpus will drive future advances in prompt-based speech generation systems.

cs.SD↗

arXiv2025

Reputation in public goods cooperation under double Q-learning protocol

Understanding and resolving cooperation dilemmas are key challenges in evolutionary game theory, which have revealed several mechanisms to address them. This paper investigates the comprehensive influence of multiple reputation-related components on public cooperation. In particular, cooperative investments in public goods game are not fixed but simultaneously depend on the reputation of group organizers and the population's cooperation willingness, hence indirectly impacting on the players' income. Additionally, individual payoff can also be directly affected by their reputation via a weighted approach which effectively evaluates the actual income of players. Unlike conventional models, the reputation change of players is non-monotonic, but may transform abruptly due to specific actions. Importantly, a theoretically supported double Q-learning algorithm is introduced to avoid overestimation bias inherent from the classical Q-learning algorithm. Our simulations reveal a significantly improved cooperation level, that is explained by a detailed Q-value analysis. We also observe the lack of massive cooperative clusters in the absence of network reciprocity. At the same time, as an intriguing phenomenon, some actors maintain moderate reputation and are continuously flipping between cooperation and defection. The robustness of our results are validated by mean-field approximation.

physics.soc-ph↗

arXiv2023

Integrating Plug-and-Play Data Priors with Weighted Prediction Error for Speech Dereverberation

Speech dereverberation aims to alleviate the detrimental effects of late-reverberant components. While the weighted prediction error (WPE) method has shown superior performance in dereverberation, there is still room for further improvement in terms of performance and robustness in complex and noisy environments. Recent research has highlighted the effectiveness of integrating physics-based and data-driven methods, enhancing the performance of various signal processing tasks while maintaining interpretability. Motivated by these advancements, this paper presents a novel dereverberation frame-work, which incorporates data-driven methods for capturing speech priors within the WPE framework. The plug-and-play strategy (PnP), specifically the regularization by denoising (RED) strategy, is utilized to incorporate speech prior information learnt from data during the optimization problem solving iterations. Experimental results validate the effectiveness of the proposed approach.

cs.SD↗

arXiv2023

AcademicGPT: Empowering Academic Research

Large Language Models (LLMs) have demonstrated exceptional capabilities across various natural language processing tasks. Yet, many of these advanced LLMs are tailored for broad, general-purpose applications. In this technical report, we introduce AcademicGPT, designed specifically to empower academic research. AcademicGPT is a continual training model derived from LLaMA2-70B. Our training corpus mainly consists of academic papers, thesis, content from some academic domain, high-quality Chinese data and others. While it may not be extensive in data scale, AcademicGPT marks our initial venture into a domain-specific GPT tailored for research area. We evaluate AcademicGPT on several established public benchmarks such as MMLU and CEval, as well as on some specialized academic benchmarks like PubMedQA, SCIEval, and our newly-created ComputerScienceQA, to demonstrate its ability from general knowledge ability, to Chinese ability, and to academic ability. Building upon AcademicGPT's foundation model, we also developed several applications catered to the academic area, including General Academic Question Answering, AI-assisted Paper Reading, Paper Review, and AI-assisted Title and Abstract Generation.

cs.CL↗

arXiv2023

CT-based Subchondral Bone Microstructural Analysis in Knee Osteoarthritis via MR-Guided Distillation Learning

Background: MR-based subchondral bone effectively predicts knee osteoarthritis. However, its clinical application is limited by the cost and time of MR. Purpose: We aim to develop a novel distillation-learning-based method named SRRD for subchondral bone microstructural analysis using easily-acquired CT images, which leverages paired MR images to enhance the CT-based analysis model during training. Materials and Methods: Knee joint images of both CT and MR modalities were collected from October 2020 to May 2021. Firstly, we developed a GAN-based generative model to transform MR images into CT images, which was used to establish the anatomical correspondence between the two modalities. Next, we obtained numerous patches of subchondral bone regions of MR images, together with their trabecular parameters (BV / TV, Tb. Th, Tb. Sp, Tb. N) from the corresponding CT image patches via regression. The distillation-learning technique was used to train the regression model and transfer MR structural information to the CT-based model. The regressed trabecular parameters were further used for knee osteoarthritis classification. Results: A total of 80 participants were evaluated. CT-based regression results of trabecular parameters achieved intra-class correlation coefficients (ICCs) of 0.804, 0.773, 0.711, and 0.622 for BV / TV, Tb. Th, Tb. Sp, and Tb. N, respectively. The use of distillation learning significantly improved the performance of the CT-based knee osteoarthritis classification method using the CNN approach, yielding an AUC score of 0.767 (95% CI, 0.681-0.853) instead of 0.658 (95% CI, 0.574-0.742) (p<.001). Conclusions: The proposed SRRD method showed high reliability and validity in MR-CT registration, regression, and knee osteoarthritis classification, indicating the feasibility of subchondral bone microstructural analysis based on CT images.

eess.IV↗

arXiv2019

Segmenting Medical MRI via Recurrent Decoding Cell

The encoder-decoder networks are commonly used in medical image segmentation due to their remarkable performance in hierarchical feature fusion. However, the expanding path for feature decoding and spatial recovery does not consider the long-term dependency when fusing feature maps from different layers, and the universal encoder-decoder network does not make full use of the multi-modality information to improve the network robustness especially for segmenting medical MRI. In this paper, we propose a novel feature fusion unit called Recurrent Decoding Cell (RDC) which leverages convolutional RNNs to memorize the long-term context information from the previous layers in the decoding phase. An encoder-decoder network, named Convolutional Recurrent Decoding Network (CRDN), is also proposed based on RDC for segmenting multi-modality medical MRI. CRDN adopts CNN backbone to encode image features and decode them hierarchically through a chain of RDCs to obtain the final high-resolution score map. The evaluation experiments on BrainWeb, MRBrainS and HVSMR datasets demonstrate that the introduction of RDC effectively improves the segmentation accuracy as well as reduces the model size, and the proposed CRDN owns its robustness to image noise and intensity non-uniformity in medical MRI.

eess.IV↗

arXiv2018

In Defense of Single-column Networks for Crowd Counting

Crowd counting usually addressed by density estimation becomes an increasingly important topic in computer vision due to its widespread applications in video surveillance, urban planning, and intelligence gathering. However, it is essentially a challenging task because of the greatly varied sizes of objects, coupled with severe occlusions and vague appearance of extremely small individuals. Existing methods heavily rely on multi-column learning architectures to extract multi-scale features, which however suffer from heavy computational cost, especially undesired for crowd counting. In this paper, we propose the single-column counting network (SCNet) for efficient crowd counting without relying on multi-column networks. SCNet consists of residual fusion modules (RFMs) for multi-scale feature extraction, a pyramid pooling module (PPM) for information fusion, and a sub-pixel convolutional module (SPCM) followed by a bilinear upsampling layer for resolution recovery. Those proposed modules enable our SCNet to fully capture multi-scale features in a compact single-column architecture and estimate high-resolution density map in an efficient way. In addition, we provide a principled paradigm for density map generation and data augmentation for training, which shows further improved performance. Extensive experiments on three benchmark datasets show that our SCNet delivers new state-of-the-art performance and surpasses previous methods by large margins, which demonstrates the great effectiveness of SCNet as a single-column network for crowd counting.

cs.CV↗

arXiv2015

Analysis of Performance of Linear Analog Codes

In this paper we carefully study the MSE performance of the linear analog codes. We have derived a lower bound of the MSE performance under Likelihood(ML) and Linear Minimal Mean Square Error(LMMSE) decoding criteria respectively. It is proved in this essay that a kind of linear analog codes called \emph {unitary codes} can simultaneously achieve both of these two bounds. At the same time, we compare the obtained linear analog codes' MSE bounds with the performance of some existing nonlinear codes. The results showed that linear analog codes are actually not very satisfying and convinced us that more concerns should be cast onto the nonlinear class in order to find powerful analog codes.

cs.IT↗

arXiv2013

Analog Turbo Codes: Turning Chaos to Reliability

Analog error correction codes, by relaxing the source space and the codeword space from discrete fields to continuous fields, present a generalization of digital codes. While linear codes are sufficient for digital codes, they are not for analog codes, and hence nonlinear mappings must be employed to fully harness the power of analog codes. This paper demonstrates new ways of building effective (nonlinear) analog codes from a special class of nonlinear, fast-diverging functions known as the chaotic functions. It is shown that the "butterfly effect" of the chaotic functions matches elegantly with the distance expansion condition required for error correction, and that the useful idea in digital turbo codes can be exploited to construct efficient turbo-like chaotic analog codes. Simulations show that the new analog codes can perform on par with, or better than, their digital counter-parts when transmitting analog sources.

cs.IT↗

arXiv2013

Iterative Decoding and Turbo Equalization: The Z-Crease Phenomenon

Iterative probabilistic inference, popularly dubbed the soft-iterative paradigm, has found great use in a wide range of communication applications, including turbo decoding and turbo equalization. The classic approach of analyzing the iterative approach inevitably use the statistical and information-theoretical tools that bear ensemble-average flavors. This paper consider the per-block error rate performance, and analyzes it using nonlinear dynamical theory. By modeling the iterative processor as a nonlinear dynamical system, we report a universal "Z-crease phenomenon:" the zig-zag or up-and-down fluctuation -- rather than the monotonic decrease -- of the per-block errors, as the number of iteration increases. Using the turbo decoder as an example, we also report several interesting motion phenomenons which were not previously reported, and which appear to correspond well with the notion of "pseudo codewords" and "stopping/trapping sets." We further propose a heuristic stopping criterion to control Z-crease and identify the best iteration. Our stopping criterion is most useful for controlling the worst-case per-block errors, and helps to significantly reduce the average-iteration numbers.

cs.IT↗

arXiv2013

Recursive-Cube-of-Rings (RCR) Revisited: Properties and Enhancement

We study recursive-cube-of-rings (RCR), a class of scalable graphs that can potentially provide rich inter-connection network topology for the emerging distributed and parallel computing infrastructure. Through rigorous proof and validating examples, we have corrected previous misunderstandings on the topological properties of these graphs, including node degree, symmetry, diameter and bisection width. To fully harness the potential of structural regularity through RCR construction, new edge connecting rules are proposed. The modified graphs, referred to as {\it Class-II RCR}, are shown to possess uniform node degrees, better connectivity and better network symmetry, and hence will find better application in parallel computing.

math.GT↗

arXiv2011

Efficient Image Transmission Through Analog Error Correction

This paper presents a new paradigm for image transmission through analog error correction codes. Conventional schemes rely on digitizing images through quantization (which inevitably causes significant bandwidth expansion) and transmitting binary bit-streams through digital error correction codes (which do not automatically differentiate the different levels of significance among the bits). To strike a better overall performance in terms of transmission efficiency and quality, we propose to use a single analog error correction code in lieu of digital quantization, digital code and digital modulation. The key is to get analog coding right. We show that this can be achieved by cleverly exploiting an elegant "butterfly" property of chaotic systems. Specifically, we demonstrate a tail-biting triple-branch baker's map code and its maximum-likelihood decoding algorithm. Simulations show that the proposed analog code can actually outperform digital turbo code, one of the best codes known to date. The results and findings discussed in this paper speak volume for the promising potential of analog codes, in spite of their rather short history.

cs.MM↗