Discovering cohesive groups is a fundamental primitive in graph-based recommender systems, underpinning tasks such as social recommendation, bundle discovery, and community-aware modeling. In interaction graphs, cohesion is often modeled as the γ-quasi-clique, an induced subgraph whose internal edge density meets a user-defined threshold γ. This formulation provides explicit control over within-group connectivity while accommodating the sparsity inherent in real-world data. However, ensuring explicit density constraints while maintaining robustness remains challenging for existing heuristic approaches. This paper presents EDQC, an effective framework for cohesive group discovery under explicit density constraints. EDQC leverages a lightweight energy diffusion process to rank vertices for localizing promising candidate regions. Guided by this ranking, the framework extracts and refines a candidate subgraph to ensure the output strictly satisfies the target density requirement. Extensive experiments on 75 real-world graphs across varying density thresholds demonstrate that EDQC identifies the largest mean γ-quasi-cliques in the vast majority of cases, achieving lower variance than the state-of-the-art methods while maintaining competitive runtime, making it a robust and practical solution for cohesive group discovery in graph-based recommender systems.

Real-time tactile perception with biological fidelity is critical for biomedical applications such as neural prosthetics and robotic surgeries, where sub-millisecond latency and micron-scale spatial resolution are essential. Event-based tactile sensors, inspired by mechanoreceptors, offer ultra-low latency and high energy efficiency but pose challenges for learning robust spatiotemporal representations under data scarcity and task variability. Current Spiking Graph Neural Networks (SGNNs) suffer from rigid spatial modeling and high computational costs, limiting deployment on edge devices. We propose MA-SGNN (Multi-view Adaptive SGNN), a lightweight brain-inspired framework emulating the biological tactile pathway: sensory encoding, feature extraction, and perceptual integration. MA-SGNN introduces: (1) a bio-hybrid spike encoder using Leaky Integrate-and- Fire neurons to capture temporal dynamics and extract biologically plausible features; (2) a multi-view adaptive graph constructor modeling structural and semantic taxel correlations via dynamic graphs; and (3) a spatiotemporal aggregator for efficient graph feature fusion. Evaluated on Ev-Objects and Ev-Containers benchmarks, MA-SGNN achieves competitive accuracy while reducing inference time by 59x and 90x versus state-of-the-art models. With only 10% training data, it maintains robust performance, dropping just 13.19%-significantly outperforming baselines. These results establish that MA-SGNN offers a biologically plausible and efficient solution for practical tactile intelligence.

Accurately identifying protein-protein interaction sites (PPIS) is a critical challenge. Existing graph neural network (GNN) methods for PPIS prediction often overlook higher-order structural patterns. We propose ComGAT-PPIS, a Community-Augmented Graph Attention Network that addresses this limitation. Our model constructs a hierarchical graph by first detecting residue communities and then applies a graph attention mechanism across this community level before fusing features back to the residue level. Experiments on standard benchmarks show ComGAT-PPIS consistently outperforms state-of-the-art models, highlighting the importance of incorporating meso-scale topology for enhancing GNN-based PPIS prediction. Our code is available at https://github.com/BiscuitZhang/ComGAT.

This paper introduces rStar, a self-play mutual reasoning approach that significantly improves reasoning capabilities of small language models (SLMs) without fine-tuning or superior models. rStar decouples reasoning into a self-play mutual generation-discrimination process. First, a target SLM augments the Monte Carlo Tree Search (MCTS) with a rich set of human-like reasoning actions to construct higher quality reasoning trajectories. Next, another SLM, with capabilities similar to the target SLM, acts as a discriminator to verify each trajectory generated by the target SLM. The mutually agreed reasoning trajectories are considered mutual consistent, thus are more likely to be correct. Extensive experiments across five SLMs demonstrate rStar can effectively solve diverse reasoning problems, including GSM8K, GSM-Hard, MATH, SVAMP, and StrategyQA. Remarkably, rStar boosts GSM8K accuracy from 12.51\% to 63.91\% for LLaMA2-7B, from 36.46\% to 81.88\% for Mistral-7B, from 74.53\% to 91.13\% for LLaMA3-8B-Instruct. Code is available at https://github.com/zhentingqi/rStar.

Vision-Language Models (VLMs) excel across diverse tasks but suffer from high inference costs in time and memory. Token sparsity mitigates inefficiencies in token usage, while neuron sparsity reduces high-dimensional computations, both offering promising solutions to enhance efficiency. Recently, these two sparsity paradigms have evolved largely in parallel, fostering the prevailing assumption that they function independently. However, a fundamental yet underexplored question remains: Do they truly operate in isolation, or is there a deeper underlying interplay that has yet to be uncovered? In this paper, we conduct the first comprehensive investigation into this question. By introducing and analyzing the matching mechanism between Core Neurons and Core Tokens, we found that key neurons and tokens for inference mutually influence and reinforce each other. Building on this insight, we propose CoreMatching, a co-adaptive sparse inference framework, which leverages the synergy between token and neuron sparsity to enhance inference efficiency. Through theoretical analysis and efficiency evaluations, we demonstrate that the proposed method surpasses state-of-the-art baselines on ten image understanding tasks and three hardware devices. Notably, on the NVIDIA Titan Xp, it achieved 5$\times$ FLOPs reduction and a 10$\times$ overall speedup. Code is released at https://github.com/wangqinsi1/2025-ICML-CoreMatching/tree/main.

Calibrating large-scale camera arrays, such as those in dome-based setups, is time-intensive and typically requires dedicated captures of known patterns. While extrinsics in such arrays are fixed due to the physical setup, intrinsics often vary across sessions due to factors like lens adjustments or temperature changes. In this paper, we propose a dense-feature-driven multi-frame calibration method that refines intrinsics directly from scene data, eliminating the necessity for additional calibration captures. Our approach enhances traditional Structure-from-Motion (SfM) pipelines by introducing an extrinsics regularization term to progressively align estimated extrinsics with ground-truth values, a dense feature reprojection term to reduce keypoint errors by minimizing reprojection loss in the feature space, and an intrinsics variance term for joint optimization across multiple frames. Experiments on the Multiface dataset show that our method achieves nearly the same precision as dedicated calibration processes, and significantly enhances intrinsics and 3D reconstruction accuracy. Fully compatible with existing SfM pipelines, our method provides an efficient and practical plug-and-play solution for large-scale camera setups. Our code is publicly available at: https://github.com/YJJfish/Multi-Cali-Anything

Multi-agent large language model (LLM) systems are increasingly adopted for complex language processing tasks that require communication and coordination among agents. However, these systems often suffer substantial overhead from repeated reprocessing of overlapping contexts across agents. In typical pipelines, once an agent receives a message from its predecessor, the full context-including prior turns-must be reprocessed from scratch, leading to inefficient processing. While key-value (KV) caching is an effective solution for avoiding redundant computation in single-agent settings where prefixes remain unchanged, it cannot be directly reused in multi-agent scenarios due to diverging prefixes introduced by agent-specific context extensions. We identify that the core challenge lies in the offset variance of KV-caches across agents. To address this, we propose **KVCOMM**, a training-free framework that enables efficient prefilling in multi-agent inference by reusing KV-caches and aligning cache offsets of overlapping contexts under diverse prefix contexts. KVCOMM estimates and adjusts KV-caches for shared content by referencing a pool of cached examples—termed *anchors*—that store observed cache deviations under varying prefixes. The anchor pool is maintained and updated online, allowing dynamic adaptation to distinct user requests and context structures. KVCOMM achieves over 70% reuse rate across diverse multi- agent workloads, including retrieval-augmented generation, math reasoning, and collaborative coding tasks, all without quality degradation. Particularly, when each fully-connected agent receives 1K input tokens with 512 prefix tokens and 512 output tokens under a five-agent setting, KVCOMM achieves up to 7.8× speedup compared to the standard prefill pipeline, reducing TTFT from ∼430ms to ∼55ms. Code is available at https://github.com/FastMAS/KVCOMM.

Increasing research suggests that microRNAs (miRNAs) serve an essential function as biomarkers in various diseases. The variations in miRNA expression can influence their corresponding mRNAs, which, in turn, regulate the expression of target genes. Recently, graph neural networks (GNNs) have been widely utilized to predict miRNA-disease associations. However, a single GNN model is insufficient for fully learning node representations. Furthermore, individual aggregation methods struggle to effectively extract diverse structural information and node weights. To address these challenges, we propose a method that incorporates Principal Neighborhood Aggregation (PNA) and Graph Attention Networks (GAT) for miRNA-disease association prediction. First, we integrated multiple datasets to construct a weighted heterogeneous graph that models miRNA-LncRNA-disease interactions. Subsequently, PNA extracted node representations using multiple aggregators simultaneously. Additionally, features derived from both PNA and GAT were fused using an attention mechanism. These combined representations were then fed into a fully connected neural network for prediction. Experimental results demonstrate that PNAGMDA achieves exceptional performance, with AUC values of 93.82% and 92.77% on HMDD v2.0 and v3.2, respectively. Case studies, along with supplementary findings, confirm PNAGMDA's reliability for miRNA-disease prediction.

General-purpose graphics processing units (GPGPUs) have become a leading platform for accelerating modern compute-intensive applications, such as large language models and generative artificial intelligence (AI). However, the lack of advanced open-source GPGPU microarchitectures has hindered high-performance research in this area. In this article, we present Ventus, a high-performance open-source GPGPU implementation built upon the RISC-V architecture with vector extension [RISC-V vector (RVV)]. Ventus introduces customized instructions and a comprehensive software toolchain to optimize performance. We deployed the design on a field programmable gate array (FPGA) platform consisting of 4 Xilinx VU19P devices, scaling up to 16 streaming multiprocessors (SMs) and supporting 256 warps. Experimental results demonstrate that Ventus exhibits key performance features comparable to commercial GPGPUs, achieving an average of 83.9% instruction reduction and 87.4% cycle per instruction (CPI) improvement over the leading open-source alternatives. Under 4-, 8-, and 16-thread configurations, Ventus maintains robust instruction per cycle (IPC) performance with values of 0.47, 0.40, and 0.32, respectively. In addition, the tensor core of Ventus attains an extra average reduction of 69.1% in instruction count and a 68.4% cycle reduction ratio when running AI-related workloads. These findings highlight Ventus as a promising solution for future high-performance GPGPU research and development, offering a robust open-source alternative to proprietary solutions. Ventus can be found onhttps://github.com/THU-DSP-LAB/ventus-gpgpu

General-purpose Graphics Processing Unit (G PG PU) has become the most popular platform for accelerating modern applications such as Large Language Models and Generative AI, while the lack of advanced open-source hardware micro architectures restricts the high-performance GPGPU research. In this work, we propose Ventus, a high-performance open-source GPGPU based on RISC- V with Vector Extension (RVV). Customized instructions and a holistic software toolchain are implemented to achieve high performance. Ventus is successfully deployed on an FPGA platform consisting of 4 Xilinx VU19P, scaling up to 16 Streaming Multiprocessors (SMs) with 256 warps. Results imply that Ventus possesses critical features of commercial GPGPUs and has achieved an average reduction of 83.9% in instruction count and 87.4% in CPI over the state-of-the-art open-source implementation. Ventus can be found on Github (https://github.com/THU-DSP-LAB/ventus-gpgpu).

Continual learning (CL) can help pre-trained vision-language models efficiently adapt to new or under-trained data distributions without re-training. Nevertheless, during the continual training of the Contrastive Language-Image Pre-training (CLIP) model, we observe that the model’s zero-shot transfer ability significantly degrades due to catastrophic forgetting. Existing CL methods can mitigate forgetting by replaying previous data. However, since the CLIP dataset is private, replay methods cannot access the pre-training dataset. In addition, replaying data of previously learned downstream tasks can enhance their performance but comes at the cost of sacrificing zero-shot performance. To address this challenge, we propose a novel method ZSCL to prevent zero-shot transfer degradation in the continual learning of vision-language models in both feature and parameter space. In the feature space, a reference dataset is introduced for distillation between the current and initial models. The reference dataset should have semantic diversity but no need to be labeled, seen in pre-training, or matched image-text pairs. In parameter space, we prevent a large parameter shift by averaging weights during the training. We propose a more challenging Multi-domain Task Incremental Learning (MTIL) benchmark to evaluate different methods, where tasks are from various domains instead of class-separated in a single dataset. Our method outperforms other methods in the traditional class-incremental learning setting and the MTIL by 9.7% average score. Our code locates at https: //github.com/Thunderbeee/ZSCL.

The interaction and dimension of points are two important axes in designing point operators to serve hierarchical 3D models. Yet, these two axes are heterogeneous and challenging to fully explore. Existing works craft point operator under a single axis and reuse the crafted operator in all parts of 3D models. This overlooks the opportunity to better combine point interactions and dimensions by exploiting varying geometry/density of 3D point clouds. In this work, we establish PIDS, a novel paradigm to jointly explore point interactions and point dimensions to serve semantic segmentation on point cloud data. We establish a large search space to jointly consider versatile point interactions and point dimensions. This supports point operators with various geometry/density considerations. The enlarged search space with heterogeneous search components calls for a better ranking of candidate models. To achieve this, we improve the search space exploration by leveraging predictor-based Neural Architecture Search (NAS), and enhance the quality of prediction by assigning unique encoding to heterogeneous search components based on their priors. We thoroughly evaluate the networks crafted by PIDS on two semantic segmentation benchmarks, showing ~ 1% mIOU improvement on SemanticKITTI and S3DIS over state-of-the-art 3D models.

Accurate and efficient on-chip power modeling is crucial to runtime power, energy, and voltage management. Such power monitoring can be achieved by designing and integrating on-chip power meters (OPMs) into the target design. In this work, we propose a new method named DEEP to automatically develop extremely efficient OPM solutions for a given design. DEEP selects OPM inputs from all individual bits in RTL signals. Such bit-level selection provides an unprecedentedly large number of input candidates and supports lower hardware cost, compared with signal-level selection in prior works. In addition, DEEP proposes a powerful two-step OPM input selection method, and it supports reporting both total power and the power of major design components. Experiments on a commercial microprocessor demonstrate that DEEP's OPM solution achieves correlation R > 0.97 in per-cycle power prediction with an unprecedented low area overhead on hardware, i.e., < 0.1% of the microprocessor layout. This reduces the OPM hardware cost by 4 -- 6× compared with the state-of-the-art solution.

With the introduction of the backdoor in deep neural networks (DNNs), much research focuses on backdoor attacks and defenses against DNNs. Since many DNN models are developed based on public datasets and pre-trained models often published by untrusted third parties, backdoors can be easily injected. The defender usually cannot access training data and does not know the target class or the triggers of the backdoor injected by the attacker. All these make it challenging to guarantee the security of decision guidance and support systems. In this paper, we proposed to detect backdoor attacks on DNNs based on model parameters analysis (MPA). We extracted and selected parameters related to the backdoor in the model's hidden layer and decision layer and trained the MPA classifier based on these parameters. We evaluated the effectiveness of the MPA classifier on various target models. The results show that the area under the receiver operating characteristic curve of the MPA classifier reaches 0.96 and 0.86 on the CIFAR10 and Troj target models, respectively. The MPA classifier improved the detection rate of backdoor attacks by 2%-6% compared with other advanced methods, with less prior knowledge and more relaxed constraints.

Increasing evidence has suggested that microRNAs (miRNAs) are important biomarkers of various diseases. Numerous graph neural network (GNN) models have been proposed for predicting miRNA-disease associations. However, the existing GNN-based methods have over-smoothing issue-the learned feature embeddings of miRNA nodes and disease nodes are indistinguishable when stacking multiple GNN layers. This issue makes the performance of the methods sensitive to the number of layers, and significantly hurts the performance when more layers are employed. In this study, we resolve this issue by a novel self-feature-based graph autoencoder model, shortened as SFGAE. The key novelty of SFGAE is to construct miRNA-self embeddings and disease-self embeddings, and let them be independent of graph interactions between two types of nodes. The novel self-feature embeddings enrich the information of typical aggregated feature embeddings, which aggregate the information from direct neighbors and hence heavily rely on graph interactions. SFGAE adopts a graph encoder with attention mechanism to concatenate aggregated feature embeddings and self-feature embeddings, and adopts a bilinear decoder to predict links. Our experiments show that SFGAE achieves state-of-the-art performance. In particular, SFGAE improves the average AUC upon recent GAEMDA [1] on the benchmark datasets HMDD v2.0 and HMDD v3.2, and consistently performs better when less (e.g. 10%) training samples are used. Furthermore, SFGAE effectively overcomes the over-smoothing issue and performs stably well on deeper models (e.g. eight layers). Finally, we carry out case studies on three human diseases, colon neoplasms, esophageal neoplasms and kidney neoplasms, and perform a survival analysis using kidney neoplasm as an example. The results suggest that SFGAE is a reliable tool for predicting potential miRNA-disease associations.

Personalized recommendation systems are widely used in many Internet services. The sparse embedding lookup in recommendation models dominates the computational cost of inference due to its intensive irregular memory accesses. Applying resistive random access memory (ReRAM) based process-in-memory (PIM) architecture to accelerate recommendation processing can avoid data movements caused by off-chip memory accesses. However, naïve adoption of ReRAM-based DNN accelerators leads to low computation parallelism and severe under-utilization of computing resources, which is caused by the fine-grained inner-product in feature interaction. In this paper, we propose Rerec, an architecture-algorithm co-designed accelerator, which specializes in fine-grained ReRAM-based inner-product engines with access-aware mapping algorithm for recommendation inference. At the architecture level, we reduce the size and increase the amount of crossbars. The crossbars are fully-connected by Analog-to-Digital Converters (ADCs) in one inner-product engine, which can adapt to the fine-grained and irregular computational patterns and improve the processing parallelism. We further explore trade-offs of (i) crossbar size vs. hardware utilization, and (ii) ADC implementation vs. area/energy efficiency to optimize the design. At the algorithm level, we propose a novel access-aware mapping (AAM) algorithm to optimize resource allocations. Our AAM algorithm tackles the problems of (i) the workload imbalance and (ii) the long recommendation inference latency induced by the great variance of access frequency of embedding vectors. Experimental results show that Rerecachieves 7.69x speedup compared with a ReRAM-based baseline design. Compared to CPU and the state-of-the-art recommendation accelerator, Rerecdemonstrates 29.26x and 3.48x performance improvement, respectively.

With the down-scaling of CMOS technology, the design complexity of very large-scale integrated is increasing. Although the application of machine learning (ML) techniques in electronic design automation (EDA) can trace its history back to the 1990s, the recent breakthrough of ML and the increasing complexity of EDA tasks have aroused more interest in incorporating ML to solve EDA tasks. In this article, we present a comprehensive review of existing ML for EDA studies, organized following the EDA hierarchy.

Compute-in-memory (CiM) is a promising method for mitigating the memory wall problem in data-intensive applications. The proposed bitwise logic-in-memory (BLiM) is targeted at data intensive applications, such as database, data encryption. This work proposes a low-power BLiM approach using the emerging nonvolatile ferroelectric FETs with direct write-back and data-adaptive dynamic sensing interface. Apart from general-purpose random-access memory, it also supports BLiM operations such as copy, not, nand, xor, and full adder (FA). The novel features of the proposed architecture include: (i) direct result-write-back based on the remnant bitline BLiM charge that avoids bitline sensing and charging operations; (ii) a fully dynamic sensing interface that needs no static reference current, but adopts data-adaptive voltage references for certain multi-operand operations, and (iii) selective bitline charging from wordline (instead of pre-charging all bitlines) to save power and also enable direct write-back. Detailed BLiM operations and benchmarking against conventional approaches show the promise of low-power computing with the FeFET-based circuit techniques.

Memristor-based logic design gives an alternative solution to improve the energy efficiency of computing systems, benefiting from combining the memory with computing units. Inspired by this thought, previous work has demonstrated various memristor-based logic families with different attributes and computation patterns. Besides, some logic synthesis tools are designed for specific memristive logic implementations. However, the poor universality and the neglect of realistic constraints in memory largely restrict the utility of these logic synthesis tools. In this paper, we propose a general logic synthesis framework for memristor-based logic design, containing a universal abstract description method for memristive logic, a mapping rules generator, and a synthesis and mapping flow. The proposed logic synthesis framework is suitable for various types of existing memristor-based logic families and takes the memory status into consideration. It is also possible to handle future memristive devices and logic families by providing the universal abstraction interface. Furthermore, we also design a circuit-partitioning-based synthesis acceleration strategy to tackle with the long synthesis time problem. Experimental results show that, our framework can generate mapping results under the restriction of limited resource, while the existing synthesis tools may fail under the same restriction, and achieve comparable synthesis results with the same resource as the existing synthesis tools, which is enough for computation and storage. And the proposed acceleration scheme can achieve ~ 1000× speedup compared with the initial one.

Graphical password (GPW) is one of various passwords used in information communication. The QR code, which is widely used in the current world, is one of GPWs. Topsnut-GPWs are new-type GPWs made by topological structures (also, called graphs) and number theory, but the existing GPWs use pictures/images almost. We design new Topsnut-GPWs by means of a graph labelling, called odd-elegant labelling. The new Topsnut-GPWs will be constructed by Topsnut-GPWs having smaller vertex numbers; in other words, they are compound Topsnut-GPWs such that they are more robust to deciphering attacks. Furthermore, the new Topsnut-GPWs can induce some mathematical problems and conjectures.