QLAM extends state-space models with quantum superposition in the hidden state for linear-time long-sequence modeling and reports consistent gains over RNN and transformer baselines on sequential image tasks.
super hub Canonical reference
Attention is all you need
Canonical reference. 89% of citing Pith papers cite this work as background.
101
Pith papers citing it
Background
89% of classified citations
hub tools
citation-role summary
background 8
method 1
citation-polarity summary
claims ledger
- background characteristics inherent in power load time series. Data-driven approaches based on artificial intelligence have become mainstream in recent years. Early methods centered on recurrent neural networks (RNNs) and convolutional neural networks (CNNs), which are adept at capturing temporal de- pendencies and inter-variable relationships [5]. With the advent of the Transformer architecture [6], attention-based models have advanced rapidly for time series forecasting, giving rise to numerous variants
- background However, these models still face challenges: their ability to explicitly model local interactions remains limited, and their interpretability is relatively weak. These drawbacks motivate our approach, which leverages physically grounded quantum walk dynamics to provide both richer local structural model- ing and improved interpretability. Formally, the self-attention mechanism in the Transformer framework [20] is defined as Attention (Q, K, V) = softmax (QKT √ d ) V(2) WhereQ, K, V∈R n×dare the
- background Generating accurate, human-like motion requires ac- counting for variability in emotion and semantic emphasis, two aspects that remain underexplored. Computational efficiency is an additional requirement for real-time robotics applications. Model architectures have evolved from recurrent networks such as long short-term memory (LSTM) [16] to attention-based transformers [17]. Adversarial and diffusion-based methods have also been proposed to improve motion realism and diver- sity [2], [14], [18]
- background Neural Machine Translation (NMT) has emerged as a pow- erful end-to-end approach for automated translation, employ- ing a single neural network to directly model the probability of a target sentence given a source sentence [1]. In recent years, NMT models have significantly improved translation quality, accompanied by a substantial expansion in model scale. Since Transformer introduced [2], the parameter count of NMT models has grown exponentially. For instance, M2M-100 (12 billion parameters) [
- background after GEMM completion while the output tiles still reside in on- chip memory (L1/L2 caches or registers), we avoid costly global memory traffic. However, conventional normalization layers operate along the feature dimension, which often misaligns with the physical data layout of GEMM outputs. To address this, we proposesBlockNorm, a normalization approach inspired by GroupNorm [71] which is originally designed to apply normalization within individual channels of a feature map. In our version of
- background bias parameters(γ, β)from conditional inputs, then modulates intermediate features viaγ⊙x+βto achieve lightweight conditional feature selection [7]. We observe that this channel- level modulation effectively adjusts feature weights with low overhead and good trainability. In contrast, attention-based cross-modal fusion typically relies on spatial weights or token- level interactions [9], [15]-[17], which increase computa- tional/parameter overhead and may complicate optimization in reinforcement
co-cited works
representative citing papers
CTQWformer fuses continuous-time quantum walks into a graph transformer and recurrent module to outperform standard GNNs and graph kernels on classification benchmarks.
CAIS delivers 1.38x end-to-end LLM training speedup over NVLS and 1.61x over T3 by making in-switch computing aware of computation memory requirements instead of treating communication as an isolated phase.
Cascaded discrete diffusion generates CAD command sequences with absorbing transitions and parameters with Gaussian, scale-invariant, and prior-preserving kernels, outperforming autoregressive and continuous diffusion baselines on the DeepCAD dataset.
Tempus delivers 607 GOPS at 10.677 W using fixed 16 AIE cores on Versal AI Edge, with 211.2x better platform-aware utility than spatial SOTA ARIES and zero URAM/DSP utilization.
Temporal autocorrelation reintroduces spectral bias in KANs for time series forecasting, which DCT preprocessing can mitigate.
Latent space probing on CogVideoX achieves 97.29% F1 for adult content detection on a new 11k-clip dataset with 4-6ms overhead.
TransSplat formulates language-driven 3D Gaussian Splatting editing as a multi-view unbalanced semantic transport problem, achieving better cross-view consistency and local editing precision than prior fusion-based methods on 8 benchmark scenes.
BiSplat-WRF applies 2D planar Gaussians rendered on angular domains plus a bilinear spatial transformer to capture electromagnetic interactions, outperforming prior NeRF and GS methods on SSIM for wireless radiance field reconstruction.
DEMUX achieves state-of-the-art multi-tab website fingerprinting accuracy by preserving boundary signals, modeling at multiple scales, and associating dispersed traffic fragments with a new three-component architecture.
STFER uses LVLM-generated identity-consistent semantic text to drive visual token filtering and expert routing for improved any-time person re-identification under clothing changes and modality shifts.
CDPR integrates polarization priors into a diffusion-based monocular depth estimator via shared latent space and adaptive gating, outperforming RGB-only methods in challenging scenes.
Creates the BGTD benchmark and mmTraffic architecture to enable explainable multimodal interpretation of encrypted network traffic using LLMs.
LiftFormer transforms monocular depth prediction into depth-oriented geometric and edge-aware subspace representations via lifting and frame theory, achieving state-of-the-art results on standard datasets.
HealthPoint represents clinical events as points in a 4D space (content, time, modality, case) and applies low-rank relational attention to achieve state-of-the-art mortality prediction from multi-level incomplete multimodal EHRs.
CBEN provides paired optical-radar images with cloud occlusion, revealing 23-33 point AP drops in clear-sky trained models and 17-29 point relative gains when models are trained on cloudy data.
D³ETOR combines debate-enhanced pseudo labeling from SAM with frequency-aware progressive debiasing in FADeNet to achieve state-of-the-art weakly-supervised camouflaged object detection using scribbles.
GT-NSGDm achieves the optimal non-asymptotic convergence rate O(1/T^{(p-1)/(3p-2)}) for decentralized nonconvex stochastic optimization under zero-mean heavy-tailed noise with p-th moment.
A full-stack LUT-as-neuron FPGA framework reports up to 205× lower latency than BNN accelerators and higher LUT efficiency than prior differentiable LUT networks at competitive binary accuracy.
AEGIS localizes sparse semantic-injecting attention heads in diffusion models and applies similarity-aware repulsion at those heads to block visual synonym jailbreaks while preserving benign generation.
MxGLUT introduces a reconfigurable LUT-centric broadcast dataflow accelerator with mixed-precision LUT-based PEs that unifies FP8-INT4 and FP8-FP8 GEMM without separate FP datapaths, reporting up to 2.16x prefill speedup and 0.492 TFLOPS/mm² area efficiency in 28nm synthesis.
Introduces the relation set completion task and RelSetE embedding model to infer missing entity-compatible relations by modeling latent patterns in observed relations, evaluated on three derived KG benchmarks.
A binary-bottleneck model discovers object and action symbols from multi-modal effect predictions on random interactions, then uses discrete planning on predicted trajectories for tabletop repositioning and stacking with few-shot generalization to novel objects.
UniT unifies online and offline 3D geometry perception via a Group Autoregressive Transformer that processes observation groups with anchor-free point map prediction and a scale-adaptive loss.
citing papers explorer
-
QLAM: A Quantum Long-Attention Memory Approach to Long-Sequence Token Modeling
QLAM extends state-space models with quantum superposition in the hidden state for linear-time long-sequence modeling and reports consistent gains over RNN and transformer baselines on sequential image tasks.
-
CTQWformer: A CTQW-based Transformer for Graph Classification
CTQWformer fuses continuous-time quantum walks into a graph transformer and recurrent module to outperform standard GNNs and graph kernels on classification benchmarks.
-
Towards Compute-Aware In-Switch Computing for LLMs Tensor-Parallelism on Multi-GPU Systems
CAIS delivers 1.38x end-to-end LLM training speedup over NVLS and 1.61x over T3 by making in-switch computing aware of computation memory requirements instead of treating communication as an isolated phase.
-
Computer-Aided Design Generation by Cascaded Discrete Diffusion Model
Cascaded discrete diffusion generates CAD command sequences with absorbing transitions and parameters with Gaussian, scale-invariant, and prior-preserving kernels, outperforming autoregressive and continuous diffusion baselines on the DeepCAD dataset.
-
Tempus: A Temporally Scalable Resource-Invariant GEMM Streaming Framework for Versal AI Edge
Tempus delivers 607 GOPS at 10.677 W using fixed 16 AIE cores on Versal AI Edge, with 211.2x better platform-aware utility than spatial SOTA ARIES and zero URAM/DSP utilization.
-
Autocorrelation Reintroduces Spectral Bias in KANs for Time Series Forecasting
Temporal autocorrelation reintroduces spectral bias in KANs for time series forecasting, which DCT preprocessing can mitigate.
-
Latent Space Probing for Adult Content Detection in Video Generative Models
Latent space probing on CogVideoX achieves 97.29% F1 for adult content detection on a new 11k-clip dataset with 4-6ms overhead.
-
RF-HiT: Rectified Flow Hierarchical Transformer for General Medical Image Segmentation
TransSplat formulates language-driven 3D Gaussian Splatting editing as a multi-view unbalanced semantic transport problem, achieving better cross-view consistency and local editing precision than prior fusion-based methods on 8 benchmark scenes.
-
Planar Gaussian Splatting with Bilinear Spatial Transformer for Wireless Radiance Field Reconstruction
BiSplat-WRF applies 2D planar Gaussians rendered on angular domains plus a bilinear spatial transformer to capture electromagnetic interactions, outperforming prior NeRF and GS methods on SSIM for wireless radiance field reconstruction.
-
DEMUX: Boundary-Aware Multi-Scale Traffic Demixing for Multi-Tab Website Fingerprinting
DEMUX achieves state-of-the-art multi-tab website fingerprinting accuracy by preserving boundary signals, modeling at multiple scales, and associating dispersed traffic fragments with a new three-component architecture.
-
Beyond Visual Cues: Semantic-Driven Token Filtering and Expert Routing for Anytime Person ReID
STFER uses LVLM-generated identity-consistent semantic text to drive visual token filtering and expert routing for improved any-time person re-identification under clothing changes and modality shifts.
-
CDPR: Cross-modal Diffusion with Polarization for Reliable Monocular Depth Estimation
CDPR integrates polarization priors into a diffusion-based monocular depth estimator via shared latent space and adaptive gating, outperforming RGB-only methods in challenging scenes.
-
Multimodal Reasoning with LLM for Encrypted Traffic Interpretation: A Benchmark
Creates the BGTD benchmark and mmTraffic architecture to enable explainable multimodal interpretation of encrypted network traffic using LLMs.
-
LiftFormer: Lifting and Frame Theory Based Monocular Depth Estimation Using Depth and Edge Oriented Subspace Representation
LiftFormer transforms monocular depth prediction into depth-oriented geometric and edge-aware subspace representations via lifting and frame theory, achieving state-of-the-art results on standard datasets.
-
A Clinical Point Cloud Paradigm for In-Hospital Mortality Prediction from Multi-Level Incomplete Multimodal EHRs
HealthPoint represents clinical events as points in a 4D space (content, time, modality, case) and applies low-rank relational attention to achieve state-of-the-art mortality prediction from multi-level incomplete multimodal EHRs.
-
CBEN -- A Multimodal Machine Learning Dataset for Cloud Robust Remote Sensing Image Understanding
CBEN provides paired optical-radar images with cloud occlusion, revealing 23-33 point AP drops in clear-sky trained models and 17-29 point relative gains when models are trained on cloudy data.
-
Debate-Enhanced Pseudo Labeling and Frequency-Aware Progressive Debiasing for Weakly-Supervised Camouflaged Object Detection with Scribble Annotations
D³ETOR combines debate-enhanced pseudo labeling from SAM with frequency-aware progressive debiasing in FADeNet to achieve state-of-the-art weakly-supervised camouflaged object detection using scribbles.
-
Decentralized Nonconvex Optimization under Heavy-Tailed Noise: Normalization and Optimal Convergence
GT-NSGDm achieves the optimal non-asymptotic convergence rate O(1/T^{(p-1)/(3p-2)}) for decentralized nonconvex stochastic optimization under zero-mean heavy-tailed noise with p-th moment.
-
FPGN: Redefining Ultra-Fast Programmable Gate-based Neural Acceleration with Differentiable LUTs
A full-stack LUT-as-neuron FPGA framework reports up to 205× lower latency than BNN accelerators and higher LUT efficiency than prior differentiable LUT networks at competitive binary accuracy.
-
AEGIS: A Mechanism-Guided Defense against Visual Synonym Jailbreaks in Text-to-Image Models
AEGIS localizes sparse semantic-injecting attention heads in diffusion models and applies similarity-aware repulsion at those heads to block visual synonym jailbreaks while preserving benign generation.
-
MxGLUT: A Reconfigurable LUT-Centric Broadcast Dataflow Accelerator for Mixed-Precision GEMM
MxGLUT introduces a reconfigurable LUT-centric broadcast dataflow accelerator with mixed-precision LUT-based PEs that unifies FP8-INT4 and FP8-FP8 GEMM without separate FP datapaths, reporting up to 2.16x prefill speedup and 0.492 TFLOPS/mm² area efficiency in 28nm synthesis.
-
Beyond Triplet Plausibility: Relation Set Completion in Knowledge Graphs
Introduces the relation set completion task and RelSetE embedding model to infer missing entity-compatible relations by modeling latent patterns in observed relations, evaluated on three derived KG benchmarks.
-
Joint Discovery of Object and Action Symbols through Effect Prediction for Robotic Manipulation Planning
A binary-bottleneck model discovers object and action symbols from multi-modal effect predictions on random interactions, then uses discrete planning on predicted trajectories for tabletop repositioning and stacking with few-shot generalization to novel objects.
-
UniT: Unified Geometry Learning with Group Autoregressive Transformer
UniT unifies online and offline 3D geometry perception via a Group Autoregressive Transformer that processes observation groups with anchor-free point map prediction and a scale-adaptive loss.
-
Memory-Augmented Query Intent Understanding for Efficient Chat-based Image Retrieval
MAQIU adds a memorization module and recall mechanism to update query intent dynamically in chat-based image retrieval, cutting FLOPs by 86.4% versus ChatIR while improving results.
-
CSI-JEPA: Towards Foundation Representations for Ubiquitous Sensing with Minimal Supervision
CSI-JEPA learns temporal-spectral representations from unlabeled CSI via masked prediction and achieves up to 10.64 percentage points accuracy gain and 98% label savings on seven real-world Wi-Fi sensing tasks.
-
Evolving Knowledge Distillation for Lightweight Neural Machine Translation
EKD trains lightweight NMT students progressively from a chain of teachers with rising capacity, achieving BLEU scores within 0.08 of the largest teacher on IWSLT-14.
-
Generating Roadside LiDAR Datasets from Vehicle-Side Datasets via Novel View Synthesis
VRS generates annotated roadside LiDAR data from vehicle observations via novel view synthesis with geometry completion and occupancy constraints, improving 3D object detection generalization.
-
Accelerating MoE with Dynamic In-Switch Computing on Multi-GPUs
DySHARP accelerates MoE expert parallelism via dynamic multimem addressing and token-centric kernel fusion to cut redundant traffic and deliver up to 1.79x speedup over prior in-switch solutions.
-
Text-to-CAD Retrieval: a Strong Baseline
Text-to-CAD retrieval is introduced as a cross-modal task with a baseline that learns joint embeddings from CAD construction sequences, point clouds, and text queries via a masked feature decoder.
-
Stage Light is Sequence$^2$: Multi-Light Control via Imitation Learning
SeqLight maps music to multi-light HSV control via SkipBART for global color prediction followed by hybrid imitation learning in a goal-conditioned MDP to decompose colors across lights.
-
RIHA: Report-Image Hierarchical Alignment for Radiology Report Generation
RIHA proposes a hierarchical alignment transformer that uses multi-scale visual and textual feature pyramids plus optimal transport to generate more accurate radiology reports from medical images.
-
DUAL-BLADE: Dual-Path NVMe-Direct KV-Cache Offloading for Edge LLM Inference
DUAL-BLADE uses a dual-path KV-cache framework with NVMe-direct access to reduce prefill and decode latency by up to 33% and 42% while improving SSD utilization 2.2x under tight memory budgets.
-
VulStyle: A Multi-Modal Pre-Training for Code Stylometry-Augmented Vulnerability Detection
VulStyle pre-trains on 4.9M functions using code, non-terminal ASTs, and stylometry features, then fine-tunes to achieve SOTA F1 gains of 4-48% on BigVul and VulDeePecker.
-
FusionCIM: Accelerating LLM Inference with Fusion-Driven Computing-in-Memory Architecture
FusionCIM is a fusion-driven CIM accelerator for LLM inference that maps QKT to IP-CIM and PV to OP-CIM, uses QO-stationary dataflow, and applies pattern-aware online softmax, delivering up to 3.86x energy savings and 1.98x speedup on LLaMA-3 at 29.4 TOPS/W.
-
BridgeACT: Bridging Human Demonstrations to Robot Actions via Unified Tool-Target Affordances
BridgeACT learns robot manipulation from human videos alone by predicting task-relevant grasp regions and 3D motion affordances that map directly to robot controllers.
-
SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference
SparKV reduces time-to-first-token by 1.3x-5.1x and energy use by 1.5x-3.3x for on-device LLM inference by adaptively choosing between cloud KV streaming and local computation while overlapping execution and adjusting for runtime conditions.
-
Lossless Compression via Chained Lightweight Neural Predictors with Information Inheritance
A new chain of lightweight neural predictors with information inheritance achieves near state-of-the-art lossless compression ratios while delivering 1.2-6.3x faster encoding and 2.8-12.3x faster decoding than PAC on GPUs.
-
Boundary-Centric Active Learning for Temporal Action Segmentation
B-ACT improves label efficiency in temporal action segmentation by selecting only boundary frames for annotation via a two-stage uncertainty-driven process that fuses neighborhood uncertainty, class ambiguity, and temporal dynamics.
-
Frequency-aware Decomposition Learning for Sensorless Wrench Forecasting on a Vibration-rich Hydraulic Manipulator
FDN uses spectral decomposition, asymmetric heads for deterministic and probabilistic wrench components, and frequency-aware filtering to forecast high-frequency wrench from proprioception, outperforming baselines on hydraulic manipulator grinding data after pretraining and transfer.
-
CODO: An Automated Compiler for Comprehensive Dataflow Optimization
CODO automates comprehensive dataflow optimization on FPGAs, achieving 1.45x-4.52x speedups on kernels and up to 33.8x on DNN models over state-of-the-art frameworks.
-
VLMaterial: Vision-Language Model-Based Camera-Radar Fusion for Physics-Grounded Material Identification
VLMaterial fuses VLMs and physics-based radar analysis via PRCA extraction and context-augmented generation to reach 96.08% material identification accuracy on 41 everyday objects without task-specific training.
-
The Salami Slicing Threat: Exploiting Cumulative Risks in LLM Systems
Salami Attack chains low-risk inputs to cumulatively trigger high-risk LLM behaviors, achieving over 90% success on GPT-4o and Gemini while resisting some defenses.
-
MAG-Net: Physics-Aware Multi-Modal Fusion of Geostationary Satellite and Radar for Severe Convective Precipitation Nowcasting
MAG-Net integrates radar dynamics with satellite IR, WV, and BTD channels via dual-stream encoding and uncertainty-weighted decoding to raise CSI40 by 0.083 over prior baselines for intense convective events.
-
Light-ResKAN: A Parameter-Sharing Lightweight KAN with Gram Polynomials for Efficient SAR Image Recognition
Light-ResKAN reaches 99.09% accuracy on MSTAR SAR images with 82.9 times fewer FLOPs and 163.78 times fewer parameters than VGG16 by combining KAN convolutions, Gram polynomials, and channel-wise parameter sharing.
-
VAN-AD: Visual Masked Autoencoder with Normalizing Flow For Time Series Anomaly Detection
A frozen ImageNet MAE plus a per-dataset normalizing flow detects time series anomalies with average AUC-ROC 0.852 over nine datasets, the best average among 15 compared baselines.
-
Rethinking Efficiency in Neural Combinatorial Optimization: Batched Preference Optimization with Mamba
ECO combines a Mamba encoder-decoder with two-stage batched DPO and LS-aware preference construction, claiming the best reported neural performance and near-linear memory scaling on TSP up to 5000 nodes and CVRP up to 1000 nodes.
-
Attention-Based Neural-Augmented Kalman Filter for Legged Robot State Estimation
AttenNKF augments InEKF with an attention-based neural compensator trained in latent space to correct foot-slip errors in legged robot state estimation.
-
BERTO: Intent-Driven Network Time Series Forecasting via Natural Language Operator Preferences
BERTO introduces a prompt-conditioned BERT framework for cellular traffic forecasting that uses a balancing loss to enable flexible trade-offs between power consumption and SLA violations using natural language inputs.
-
SpO$_2$ Predictor-Guided Stage-Wise Time-Frequency Reconstruction of Low-Quality Dual-Wavelength PPG for Oxygen Saturation Estimation
SpO2-predictor-guided masked time-frequency PPG reconstruction yields subject-level MAE of 2.882% (OpenOximetry) and 2.359% (private wearable) versus stronger baselines.