Yilong Chen, Guangxu Zhu, Jie Xu
This paper investigates the effect of imperfect channel state information (CSI) on the over-the-air computation (AirComp) system, in which multiple wireless devices (WDs) send individual messages to one access point (AP) for distributed functional computation. By particularly considering the channel estimation errors, we jointly design the transmit coefficients at the WDs and the receive strategy at the AP, for minimizing the computation mean squared error (MSE). First, we consider the single-input single-output (SISO) case with each WD and AP equipped with one single antenna, in which the globally optimal solution to the computation MSE minimization problem is obtained in closed form. Next, we consider the single-input multiple-output (SIMO) case with multiple receive antennas at the AP, in which a high-quality solution is obtained based on alternating optimization and convex optimization. For both cases, the optimized power control solution at the WDs follows a threshold-based regularized channel inversion structure; while for the SIMO case, the receive beamforming at the AP follows a sum-minimum MSE (MMSE) structure. It is shown that with finite receive antennas, a non-zero computation MSE is inevitable due to the channel estimation errors even when the WDs' transmit powers become infinity; while with massive receive antennas, a vanishing MSE is achievable when the channel vectors are independent and identically distributed. Finally, numerical results are provided to demonstrate the effectiveness of the proposed designs.
Yilong Chen, Huijun Xing, Jie Xu, Lexi Xu, Shuguang Cui
This paper studies the over-the-air computation (AirComp) in an orthogonal frequency division multiplexing (OFDM) system with imperfect channel state information (CSI), in which multiple single-antenna wireless devices (WDs) simultaneously send uncoded signals to a multi-antenna access point (AP) for distributed functional computation over multiple subcarriers. In particular, we consider two scenarios with best-effort and error-constrained computation tasks, with the objectives of minimizing the average computation mean squared error (MSE) and the computation outage probability over the multiple subcarriers, respectively. Towards this end, we jointly optimize the transmit coefficients at the WDs and the receive beamforming vectors at the AP over subcarriers, subject to the maximum transmit power constraints at individual WDs. First, for the special case with a single receive antenna at the AP, we propose the semi-closed-form globally optimal solutions to the two problems using the Lagrange-duality method. It is shown that at each subcarrier, the WDs' optimized power control policy for average MSE minimization follows a regularized channel inversion structure, while that for computation outage probability minimization follows an on-off regularized channel inversion, with the regularization dependent on the transmit power budget and channel estimation error. Next, for the general case with multiple receive antennas at the AP, we present efficient algorithms based on alternating optimization and convex optimization to find converged solutions to both problems.
Yilong Chen, Linhao Zhang, Junyuan Shang, Zhenyu Zhang, Tingwen Liu, Shuohuan Wang, Yu Sun
Large language models (LLMs) with billions of parameters demonstrate impressive performance. However, the widely used Multi-Head Attention (MHA) in LLMs incurs substantial computational and memory costs during inference. While some efforts have optimized attention mechanisms by pruning heads or sharing parameters among heads, these methods often lead to performance degradation or necessitate substantial continued pre-training costs to restore performance. Based on the analysis of attention redundancy, we design a Decoupled-Head Attention (DHA) mechanism. DHA adaptively configures group sharing for key heads and value heads across various layers, achieving a better balance between performance and efficiency. Inspired by the observation of clustering similar heads, we propose to progressively transform the MHA checkpoint into the DHA model through linear fusion of similar head parameters step by step, retaining the parametric knowledge of the MHA checkpoint. We construct DHA models by transforming various scales of MHA checkpoints given target head budgets. Our experiments show that DHA remarkably requires a mere 0.25\% of the original model's pre-training budgets to achieve 97.6\% of performance while saving 75\% of KV cache. Compared to Group-Query Attention (GQA), DHA achieves a 5$\times$ training acceleration, a maximum of 13.93\% performance improvement under 0.01\% pre-training budget, and 4\% relative improvement under 0.05\% pre-training budget.
Yilong Chen, Naibin Gu, Junyuan Shang, Zhenyu Zhang, Yuchen Feng, Jiawei Sheng, Tingwen Liu, Shuohuan Wang, Yu Sun, Hua Wu, Haifeng Wang
Mixture-of-Experts (MoE) decouples model capacity from per-token computation, yet their scalability remains limited by the physical dimensions of depth and width. To overcome this, we propose Mixture of Universal Experts (MOUE),a MoE generalization introducing a novel scaling dimension: Virtual Width. In general, MoUE aims to reuse a universal layer-agnostic expert pool across layers, converting depth into virtual width under a fixed per-token activation budget. However, two challenges remain: a routing path explosion from recursive expert reuse, and a mismatch between the exposure induced by reuse and the conventional load-balancing objectives. We address these with three core components: a Staggered Rotational Topology for structured expert sharing, a Universal Expert Load Balance for depth-aware exposure correction, and a Universal Router with lightweight trajectory state for coherent multi-step routing. Empirically, MoUE consistently outperforms matched MoE baselines by up to 1.3% across scaling regimes, enables progressive conversion of existing MoE checkpoints with up to 4.2% gains, and reveals a new scaling dimension for MoE architectures.
Zhixiong Nan, Yilong Chen, Tianfei Zhou, Tao Xiang
This paper addresses the problem of on-road object importance estimation, which utilizes video sequences captured from the driver's perspective as the input. Although this problem is significant for safer and smarter driving systems, the exploration of this problem remains limited. On one hand, publicly-available large-scale datasets are scarce in the community. To address this dilemma, this paper contributes a new large-scale dataset named Traffic Object Importance (TOI). On the other hand, existing methods often only consider either bottom-up feature or single-fold guidance, leading to limitations in handling highly dynamic and diverse traffic scenarios. Different from existing methods, this paper proposes a model that integrates multi-fold top-down guidance with the bottom-up feature. Specifically, three kinds of top-down guidance factors (ie, driver intention, semantic context, and traffic rule) are integrated into our model. These factors are important for object importance estimation, but none of the existing methods simultaneously consider them. To our knowledge, this paper proposes the first on-road object importance estimation model that fuses multi-fold top-down guidance factors with bottom-up feature. Extensive experiments demonstrate that our model outperforms state-of-the-art methods by large margins, achieving 23.1% Average Precision (AP) improvement compared with the recently proposed model (ie, Goal).
Yilong Chen, Chao Hu, Zixiang Ren, Han Hu, Jie Xu, Lexi Xu, Lei Liu, Shuguang Cui
This paper considers a multi-functional orthogonal frequency division multiplexing (OFDM) system with integrated sensing, communication, and powering (ISCAP), in which a multi-antenna base station (BS) transmits OFDM signals to simultaneously deliver information to multiple information receivers (IRs), provide energy supply to multiple energy receivers (ERs), and sense potential targets based on the echo signals. To facilitate ISCAP, the BS employs the joint transmit beamforming design by sending dedicated sensing/energy beams jointly with information beams. Furthermore, we consider the beam scanning for sensing, in which the joint beams scan in different directions over time to sense potential targets. In order to ensure the sensing beam scanning performance and meet the communication and powering requirements, it is essential to properly schedule IRs and ERs and design the resource allocation over time, frequency, and space. More specifically, we optimize the joint transmit beamforming over multiple OFDM symbols and subcarriers, with the objective of minimizing the average beampattern matching error of beam scanning for sensing, subject to the constraints on the average communication rates at IRs and the average harvested power at ERs. We find converged high-quality solutions to the formulated problem by proposing efficient iterative algorithms based on advanced optimization techniques. We also develop various heuristic designs based on the principles of zero-forcing (ZF) beamforming, round-robin user scheduling, and time switching, respectively. Numerical results show that our proposed algorithms adaptively generate information and sensing/energy beams at each time-frequency slot to match the scheduled IRs/ERs with the desired scanning beam, significantly outperforming the heuristic designs.
Yilong Chen, Junyuan Shang, Zhengyu Zhang, Jiawei Sheng, Tingwen Liu, Shuohuan Wang, Yu Sun, Hua Wu, Haifeng Wang
Transformer models encounter challenges in scaling hidden dimensions efficiently, as uniformly increasing them inflates computational and memory costs while failing to emphasize the most relevant features for each token. For further understanding, we study hidden dimension sparsity and observe that trained Transformers utilize only a small fraction of token dimensions, revealing an "activation flow" pattern. Notably, there are shared sub-dimensions with sustained activation across multiple consecutive tokens and specialized sub-dimensions uniquely activated for each token. To better model token-relevant sub-dimensions, we propose MoHD (Mixture of Hidden Dimensions), a sparse conditional activation architecture. Particularly, MoHD employs shared sub-dimensions for common token features and a routing mechanism to dynamically activate specialized sub-dimensions. To mitigate potential information loss from sparsity, we design activation scaling and group fusion mechanisms to preserve activation flow. In this way, MoHD expands hidden dimensions with negligible increases in computation or parameters, efficient training and inference while maintaining performance. Evaluations across 10 NLP tasks show that MoHD surpasses Vanilla Transformers in parameter efficiency and task performance. It achieves 1.7% higher performance with 50% fewer activation parameters and 3.7% higher performance with a 3x parameter expansion at constant activation cost. MOHD offers a new perspective for scaling the model, showcasing the potential of hidden dimension sparsity to boost efficiency
Yilong Chen, Xiang Bai, Zhibin Wang, Chengyu Bai, Yuhan Dai, Ming Lu, Shanghang Zhang
Video Large Language Models (Video-LLMs) have demonstrated significant potential in the areas of video captioning, search, and summarization. However, current Video-LLMs still face challenges with long real-world videos. Recent methods have introduced a retrieval mechanism that retrieves query-relevant KV caches for question answering, enhancing the efficiency and accuracy of long real-world videos. However, the compression and retrieval of KV caches are still not fully explored. In this paper, we propose \textbf{StreamKV}, a training-free framework that seamlessly equips Video-LLMs with advanced KV cache retrieval and compression. Compared to previous methods that used uniform partitioning, StreamKV dynamically partitions video streams into semantic segments, which better preserves semantic information. For KV cache retrieval, StreamKV calculates a summary vector for each segment to retain segment-level information essential for retrieval. For KV cache compression, StreamKV introduces a guidance prompt designed to capture the key semantic elements within each segment, ensuring only the most informative KV caches are retained for answering questions. Moreover, StreamKV unifies KV cache retrieval and compression within a single module, performing both in a layer-adaptive manner, thereby further improving the effectiveness of streaming video question answering. Extensive experiments on public StreamingVQA benchmarks demonstrate that StreamKV significantly outperforms existing Online Video-LLMs, achieving superior accuracy while substantially improving both memory efficiency and computational latency. The code has been released at https://github.com/sou1p0wer/StreamKV.
Pengwei Wang, Yilong Chen, Wan Su, Jie Wang, Teng Ma, Haoyong Yu
Deep learning models have become a powerful tool in knee angle estimation for lower limb prostheses, owing to their adaptability across various gait phases and locomotion modes. Current methods utilize Multi-Layer Perceptrons (MLP), Long-Short Term Memory Networks (LSTM), and Convolutional Neural Networks (CNN), predominantly analyzing motion information from the thigh. Contrary to these approaches, our study introduces a holistic perspective by integrating whole-body movements as inputs. We propose a transformer-based probabilistic framework, termed the Angle Estimation Probabilistic Model (AEPM), that offers precise angle estimations across extensive scenarios beyond walking. AEPM achieves an overall RMSE of 6.70 degrees, with an RMSE of 3.45 degrees in walking scenarios. Compared to the state of the art, AEPM has improved the prediction accuracy for walking by 11.31%. Our method can achieve seamless adaptation between different locomotion modes. Also, this model can be utilized to analyze the synergy between the knee and other joints. We reveal that the whole body movement has valuable information for knee movement, which can provide insights into designing sensors for prostheses. The code is available at https://github.com/penway/Beyond-Gait-AEPM.
Yilong Chen, Zixiang Ren, Xianghao Yu, Lei Liu, Jie Xu
This letter exploits moving arrays to enable nearfield multiple-input multiple-output (MIMO) sensing via a limited number of antenna elements. We consider a scenario where a base station (BS) is equipped with a uniform linear array (ULA) on a moving platform. The objective is to locate a point target in the two-dimensional (2D) space by leveraging the near-field channel characteristics created by the movement of antenna arrays. Under this setup, we analyze the Cramer-Rao bound (CRB) for estimating the target's 2D coordinate, which provides the fundamental sensing performance limits for localization. It is revealed that our proposed design with a moving array achieves a CRB that is proportional to the CRB obtained by an equivalent extremely large ULA matching the platform's size. This shows that the movement of antenna array significantly enlarges its effective aperture to enable near-field sensing. Numerical results show that the proposed moving array design substantially enhances the target estimation performance compared to the conventional fixed array benchmark.
Yilong Chen, Guoxia Wang, Junyuan Shang, Shiyao Cui, Zhenyu Zhang, Tingwen Liu, Shuohuan Wang, Yu Sun, Dianhai Yu, Hua Wu
Large Language Models (LLMs) have ignited an innovative surge of AI applications, marking a new era of exciting possibilities equipped with extended context windows. However, hosting these models is cost-prohibitive mainly due to the extensive memory consumption of KV Cache involving long-context modeling. Despite several works proposing to evict unnecessary tokens from the KV Cache, most of them rely on the biased local statistics of accumulated attention scores and report performance using unconvincing metric like perplexity on inadequate short-text evaluation. In this paper, we propose NACL, a general framework for long-context KV cache eviction that achieves more optimal and efficient eviction in a single operation during the encoding phase. Due to NACL's efficiency, we combine more accurate attention score statistics in PROXY TOKENS EVICTION with the diversified random eviction strategy of RANDOM EVICTION, aiming to alleviate the issue of attention bias and enhance the robustness in maintaining pivotal tokens for long-context modeling tasks. Notably, our method significantly improves the performance on short- and long-text tasks by 80% and 76% respectively, reducing KV Cache by up to 50% with over 95% performance maintenance. The code is available at https://github.com/PaddlePaddle/Research/tree/master/NLP/ACL2024-NACL.
Yilong Chen, Haocheng Hua, Jie Xu, Derrick Wing Kwan Ng
This paper unifies integrated sensing and communication (ISAC) and simultaneous wireless information and power transfer (SWIPT), by investigating a new multi-functional multiple-input multiple-output (MIMO) system integrating wireless sensing, communication, and powering. In this system, one multi-antenna hybrid access point (H-AP) transmits wireless signals to communicate with one multi-antenna information decoding (ID) receiver, wirelessly charge one multi-antenna energy harvesting (EH) receiver, and perform radar target sensing based on the echo signal at the same time. Under this setup, we aim to reveal the fundamental performance tradeoff limits among sensing, communication, and powering, in terms of the estimation Cramer-Rao bound (CRB), achievable communication rate, and harvested energy level, respectively. In particular, we consider two different target models for radar sensing, namely the point and extended targets, for which we are interested in estimating the target angle and the complete target response matrix, respectively. For both models, we define the achievable CRB-rate-energy (C-R-E) region and characterize its Pareto boundary by maximizing the achievable rate at the ID receiver, subject to the estimation CRB requirement for target sensing, the harvested energy requirement at the EH receiver, and the maximum transmit power constraint at the H-AP. We obtain the well-structured optimal transmit covariance solutions to the two formulated problems by applying advanced convex optimization techniques. Numerical results show the optimal C-R-E region boundary achieved by our proposed design, as compared to the benchmark schemes based on time switching and eigenmode transmission (EMT).
Yilong Chen, Zhixiong Nan, Tao Xiang
The problem of predicting driver attention from the driving perspective is gaining increasing research focus due to its remarkable significance for autonomous driving and assisted driving systems. The driving experience is extremely important for safe driving,a skilled driver is able to effortlessly predict oncoming danger (before it becomes salient) based on the driving experience and quickly pay attention to the corresponding zones. However, the nonobjective driving experience is difficult to model, so a mechanism simulating the driver experience accumulation procedure is absent in existing methods, and the current methods usually follow the technique line of saliency prediction methods to predict driver attention. In this paper, we propose a FeedBack Loop Network (FBLNet), which attempts to model the driving experience accumulation procedure. By over-and-over iterations, FBLNet generates the incremental knowledge that carries rich historically-accumulative and long-term temporal information. The incremental knowledge in our model is like the driving experience of humans. Under the guidance of the incremental knowledge, our model fuses the CNN feature and Transformer feature that are extracted from the input image to predict driver attention. Our model exhibits a solid advantage over existing methods, achieving an outstanding performance improvement on two driver attention benchmark datasets.
Yilong Chen, Haocheng Hua, Jie Xu
This paper unifies integrated sensing and communication (ISAC) and simultaneous wireless information and power transfer (SWIPT), by investigating a new multi-functional multiple-input multiple-output (MIMO) system integrating wireless sensing, communication, and powering. In this system, one multi-antenna hybrid access point (H-AP) transmits wireless signals to communicate with one multi-antenna information decoding (ID) receiver, wirelessly charge one multi-antenna energy harvesting (EH) receiver, and perform radar sensing for a point target based on the echo signal at the same time. Under this setup, we aim to reveal the fundamental performance tradeoff limits of sensing, communication, and powering, in terms of the estimation Cram{é}r-Rao bound (CRB), achievable communication rate, and harvested energy level, respectively. Towards this end, we define the achievable CRB-rate-energy (C-R-E) region and characterize its Pareto boundary by maximizing the achievable rate at the ID receiver, subject to the estimation CRB requirement for target sensing, the harvested energy requirement at the EH receiver, and the maximum transmit power constraint at the H-AP. We obtain the semi-closed-form optimal transmit covariance solution to the formulated problem by applying advanced convex optimization techniques. Numerical results show the optimal C-R-E region boundary achieved by our proposed design, as compared to the benchmark scheme based on time switching.
Yilong Chen, Zixiang Ren, Jie Xu, Rui Zhang
This paper studies the intelligent reflecting surface (IRS) deployment optimization problem for IRS-enabled integrated sensing and communications (ISAC) systems, in which multiple IRSs are strategically deployed at candidate locations to assist a base station (BS) to enhance the coverage of both sensing and communications. We present an environment-aware IRS deployment design via exploiting the channel knowledge map (CKM), which provides the channel state information (CSI) between each candidate IRS location and BS or targeted sensing/communication points. Based on the obtained CSI from CKM, we optimize the deployment of IRSs, jointly with the BS's transmit beamforming and IRSs' reflective beamforming during operation, with the objective of minimizing the system cost, while guaranteeing the minimum illumination power requirements at sensing areas and the minimum signal-to-noise ratio (SNR) requirements at communication areas. In particular, we consider two cases when the IRSs' reflective beamforming optimization can be implemented dynamically in real time and quasi-stationarily over the whole operation period, respectively. For both cases, the joint IRS deployment and transmit/reflective beamforming designs are formulated as mixed-integer non-convex optimization problems, which are solved via the successive convex approximation (SCA)-based relax-and-bound method. Specifically, we first relax the binary IRS deployment indicators into continuous variables, then find converged solutions via SCA, and finally round relaxed indicators back to binary values. Numerical results demonstrate the effectiveness of our proposed algorithms in reducing the system cost while meeting the sensing and communication requirements.
Chuanyu Tang, Yilong Chen, Zhenyu Zhang, Junyuan Shang, Wenyuan Zhang, Yong Huang, Tingwen Liu
Low-Rank Adaptation (LoRA) drives research to align its performance with full fine-tuning. However, significant challenges remain: (1) Simply increasing the rank size of LoRA does not effectively capture high-rank information, which leads to a performance bottleneck.(2) MoE-style LoRA methods substantially increase parameters and inference latency, contradicting the goals of efficient fine-tuning and ease of application. To address these challenges, we introduce Mixture of Ranks (MoR), which learns rank-specific information for different tasks based on input and efficiently integrates multi-rank information. We firstly propose a new framework that equates the integration of multiple LoRAs to expanding the rank of LoRA. Moreover, we hypothesize that low-rank LoRA already captures sufficient intrinsic information, and MoR can derive high-rank information through mathematical transformations of the low-rank components. Thus, MoR can reduces the learning difficulty of LoRA and enhances its multi-task capabilities. MoR achieves impressive results, with MoR delivering a 1.31\% performance improvement while using only 93.93\% of the parameters compared to baseline methods.
Yilong Chen, Junyuan Shang, Zhenyu Zhang, Yanxi Xie, Jiawei Sheng, Tingwen Liu, Shuohuan Wang, Yu Sun, Hua Wu, Haifeng Wang
Large language models (LLMs) face inherent performance bottlenecks under parameter constraints, particularly in processing critical tokens that demand complex reasoning. Empirical analysis reveals challenging tokens induce abrupt gradient spikes across layers, exposing architectural stress points in standard Transformers. Building on this insight, we propose Inner Thinking Transformer (ITT), which reimagines layer computations as implicit thinking steps. ITT dynamically allocates computation through Adaptive Token Routing, iteratively refines representations via Residual Thinking Connections, and distinguishes reasoning phases using Thinking Step Encoding. ITT enables deeper processing of critical tokens without parameter expansion. Evaluations across 162M-466M parameter models show ITT achieves 96.5\% performance of a 466M Transformer using only 162M parameters, reduces training data by 43.2\%, and outperforms Transformer/Loop variants in 11 benchmarks. By enabling elastic computation allocation during inference, ITT balances performance and efficiency through architecture-aware optimization of implicit thinking pathways.
Yilong Chen, Yanxi Xie, Zitian Gao, He Xin, Yihao Xiao, Renbiao Liu, Haoming Luo, Yifan Luo, Zhengmao Ye, Tingwen Liu, Xin Zhao, Ran Tao, Bryan Dai
Large token-indexed lookup tables provide a compute-decoupled scaling path, but their practical gains are often limited by poor parameter efficiency and rapid memory growth. We attribute these limitations to Zipfian under-training of the long tail, heterogeneous demand across layers, and "slot collapse" that produces redundant embeddings. To address this, we propose X-GRAM, a frequency-aware dynamic token-injection framework. X-GRAM employs hybrid hashing and alias mixing to compress the tail while preserving head capacity, and refines retrieved vectors via normalized SwiGLU ShortConv to extract diverse local n-gram features. These signals are integrated into attention value streams and inter-layer residuals using depth-aware gating, effectively aligning static memory with dynamic context. This design introduces a memory-centric scaling axis that decouples model capacity from FLOPs. Extensive evaluations at the 0.73B and 1.15B scales show that X-GRAM improves average accuracy by as much as 4.4 points over the vanilla backbone and 3.2 points over strong retrieval baselines, while using substantially smaller tables in the 50% configuration. Overall, by decoupling capacity from compute through efficient memory management, X-GRAM offers a scalable and practical paradigm for future memory-augmented architectures. Code aviliable in https://github.com/Longyichen/X-gram.
Zhou Zhang, Dongzeng Tan, Jiaan Wang, Yilong Chen, Jiarong Xu
Emojis have gained immense popularity on social platforms, serving as a common means to supplement or replace text. However, existing data mining approaches generally either completely ignore or simply treat emojis as ordinary Unicode characters, which may limit the model's ability to grasp the rich semantic information in emojis and the interaction between emojis and texts. Thus, it is necessary to release the emoji's power in social media data mining. To this end, we first construct a heterogeneous graph consisting of three types of nodes, i.e. post, word and emoji nodes to improve the representation of different elements in posts. The edges are also well-defined to model how these three elements interact with each other. To facilitate the sharing of information among post, word and emoji nodes, we propose a graph pre-train framework for text and emoji co-modeling, which contains two graph pre-training tasks: node-level graph contrastive learning and edge-level link reconstruction learning. Extensive experiments on the Xiaohongshu and Twitter datasets with two types of downstream tasks demonstrate that our approach proves significant improvement over previous strong baseline methods.
Zongyi Xu, Zhongpeng Lang, Yilong Chen, Shanshan Zhao, Xiaoshui Huang, Yifan Zuo, Yan Zhang, Qianni Zhang, Xinbo Gao
Cross-source point cloud registration, which aims to align point cloud data from different sensors, is a fundamental task in 3D vision. However, compared to the same-source point cloud registration, cross-source registration faces two core challenges: the lack of publicly available large-scale real-world datasets for training the deep registration models, and the inherent differences in point clouds captured by multiple sensors. The diverse patterns induced by the sensors pose great challenges in robust and accurate point cloud feature extraction and matching, which negatively influence the registration accuracy. To advance research in this field, we construct Cross3DReg, the currently largest and real-world multi-modal cross-source point cloud registration dataset, which is collected by a rotating mechanical lidar and a hybrid semi-solid-state lidar, respectively. Moreover, we design an overlap-based cross-source registration framework, which utilizes unaligned images to predict the overlapping region between source and target point clouds, effectively filtering out redundant points in the irrelevant regions and significantly mitigating the interference caused by noise in non-overlapping areas. Then, a visual-geometric attention guided matching module is proposed to enhance the consistency of cross-source point cloud features by fusing image and geometric information to establish reliable correspondences and ultimately achieve accurate and robust registration. Extensive experiments show that our method achieves state-of-the-art registration performance. Our framework reduces the relative rotation error (RRE) and relative translation error (RTE) by $63.2\%$ and $40.2\%$, respectively, and improves the registration recall (RR) by $5.4\%$, which validates its effectiveness in achieving accurate cross-source registration.