{"total":88,"rows":[{"id":"2606.19348","title":"DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence","created":"2026-04-26","cited_by_count":1040,"area":"moe"},{"id":"2603.15031","title":"Attention Residuals","created":"2026-03-16","cited_by_count":65,"area":null},{"id":"2608.30320","title":"On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability","created":"2026-08-31","cited_by_count":35,"area":"moe"},{"id":"2601.21579","title":"KromHC: Manifold-Constrained Hyper-Connections with Kronecker-Product Residual Matrices","created":"2026-01-29","cited_by_count":11,"area":null},{"id":"2602.08064","title":"SiameseNorm: Breaking the Barrier to Reconciling Pre/Post-Norm","created":"2026-02-08","cited_by_count":10,"area":null},{"id":"2601.00417","title":"Deep Delta Learning","created":"2026-01-01","cited_by_count":9,"area":null},{"id":"2603.15619","title":"Mixture-of-Depths Attention","created":"2026-03-16","cited_by_count":6,"area":null},{"id":"2602.09009","title":"ANCRe: Adaptive Neural Connection Reassignment for Efficient Depth Scaling","created":"2026-02-09","cited_by_count":5,"area":null},{"id":"2602.18308","title":"JPmHC Dynamical Isometry via Orthogonal Hyper-Connections","created":"2026-02-20","cited_by_count":5,"area":null},{"id":"2602.05943","title":"Orthogonal Model Merging","created":"2026-02-05","cited_by_count":4,"area":null},{"id":"2603.20896","title":"Spectral-Sphere-Constrained Hyper-Connections","created":"2026-03-21","cited_by_count":4,"area":null},{"id":"2605.12374","title":"Fill the GAP: A Granular Alignment Paradigm for Visual Reasoning in Multimodal Large Language Models","created":"2026-05-12","cited_by_count":4,"area":"lr"},{"id":"2606.03483","title":"Analyzing Stream Collapse in Hyper-Connections: From Diagnosis to Mitigation","created":"2026-06-02","cited_by_count":4,"area":null},{"id":"2607.14530","title":"xHC: Expanded Hyper-Connections","created":"2026-07-16","cited_by_count":4,"area":"moe"},{"id":"2601.21686","title":"Don't be so Stief! Learning KV Cache low-rank approximation over the Stiefel manifold","created":"2026-01-29","cited_by_count":3,"area":"inference"},{"id":"2602.08306","title":"TextResNet: Decoupling and Routing Optimization Signals in Compound AI Systems via Deep Residual Tuning","created":"2026-02-09","cited_by_count":3,"area":null},{"id":"2603.10846","title":"Towards Cold-Start Drafting and Continual Refining: A Value-Driven Memory Approach with Application to NPU Kernel Synthesis","created":"2026-03-11","cited_by_count":3,"area":null},{"id":"2604.05629","title":"A Unified Foundation Model for All-in-One Multi-Modal Remote Sensing Image Restoration and Fusion with Language Prompting","created":"2026-04-07","cited_by_count":3,"area":null},{"id":"2605.11172","title":"Optimistic Dual Averaging Unifies Modern Optimizers","created":"2026-05-11","cited_by_count":3,"area":null},{"id":"2605.21724","title":"TBP-mHC: full expressivity for manifold-constrained hyper connections through transportation polytopes","created":"2026-05-20","cited_by_count":3,"area":null},{"id":"2606.07574","title":"Accelerating Birkhoff Projection for Manifold-Constrained Hyper-Connections","created":"2026-05-26","cited_by_count":3,"area":null},{"id":"2606.14255","title":"ReactVLA: Fast and Lightweight Reactive Robot Manipulation via Improved Mean Flow Action Generation","created":"2026-06-12","cited_by_count":3,"area":"vla"},{"id":"2607.28611","title":"Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers","created":"2026-07-30","cited_by_count":3,"area":"diffusion"},{"id":"2602.01212","title":"SimpleGPT: Improving GPT via A Simple Normalization Strategy","created":"2026-02-01","cited_by_count":2,"area":null},{"id":"2603.13381","title":"Beyond Linearity in Attention Projections: The Case for Nonlinear Queries","created":"2026-03-11","cited_by_count":2,"area":null},{"id":"2604.13556","title":"YOCO++: Enhancing YOCO with KV Residual Connections for Efficient LLM Inference","created":"2026-04-15","cited_by_count":2,"area":"inference"},{"id":"2605.20708","title":"Rethinking Cross-Layer Information Routing in Diffusion Transformers","created":"2026-05-20","cited_by_count":2,"area":"diffusion"},{"id":"2606.28116","title":"Mechanism-Driven Monitors for Preemptive Detection of LLM Training Instability","created":"2026-06-26","cited_by_count":2,"area":null},{"id":"2609.05309","title":"How Does mHC Use Its Residual Streams? Selective Routing and Near-Identity Mixing","created":"2026-09-04","cited_by_count":2,"area":null},{"id":"2601.23039","title":"Avoiding Premature Collapse: Adaptive Annealing for Entropy-Regularized Structural Inference","created":"2026-01-30","cited_by_count":1,"area":null},{"id":"2602.01763","title":"A Provable Expressiveness Hierarchy in Hybrid Linear-Full Attention","created":"2026-02-02","cited_by_count":1,"area":null},{"id":"2602.06471","title":"Revisiting the Shape Convention of Transformer Language Models","created":"2026-02-06","cited_by_count":1,"area":"inference"},{"id":"2603.05369","title":"Progressive Residual Warmup for Language Model Pretraining","created":"2026-03-05","cited_by_count":1,"area":null},{"id":"2604.23036","title":"Preserving Long-Tailed Expert Information in Mixture-of-Experts Tuning","created":"2026-04-24","cited_by_count":1,"area":"moe"},{"id":"2605.03953","title":"Transformers with Selective Access to Early Representations","created":"2026-05-05","cited_by_count":1,"area":null},{"id":"2605.04421","title":"FLUID: Continuous-Time Hyperconnected Sparse Transformer for Sink-Free Learning","created":"2026-05-06","cited_by_count":1,"area":null},{"id":"2605.17842","title":"SNLP: Layer-Parallel Inference via Structured Newton Corrections","created":"2026-05-18","cited_by_count":1,"area":null},{"id":"2605.23259","title":"Multi-Gate Residuals","created":"2026-05-22","cited_by_count":1,"area":null},{"id":"2606.06470","title":"PC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-Training","created":"2026-06-04","cited_by_count":1,"area":null},{"id":"2606.23235","title":"A First-Order Mean Field Control Analysis of Transformer Layers under Cross-Entropy Training","created":"2026-06-22","cited_by_count":1,"area":null},{"id":"2607.09694","title":"Low-Rank Attention Residuals","created":"2026-06-19","cited_by_count":1,"area":null},{"id":"2607.20145","title":"SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD","created":"2026-07-22","cited_by_count":1,"area":"moe"},{"id":"2609.02672","title":"oHC: Orthogonal Hyper-Connections on SO(4) via Quaternions","created":"2026-09-02","cited_by_count":1,"area":null},{"id":"2609.26368","title":"HySparse2: Hybrid Sparse Attention with Two-Level KV Sharing","created":"2026-09-22","cited_by_count":1,"area":"inference"},{"id":"2609.32534","title":"DepthBench: Measuring How Residual Connections Enable More Computational Depth","created":"2026-09-26","cited_by_count":1,"area":null},{"id":"2601.01014","title":"Geometric and Dynamic Scaling in Deep Transformers","created":"2026-01-03","cited_by_count":0,"area":null},{"id":"2601.15757","title":"White-Box mHC: Electromagnetic Spectrum-Aware and Interpretable Stream Interactions for Hyperspectral Image Classification","created":"2026-01-22","cited_by_count":0,"area":null},{"id":"2603.03418","title":"mHC-HSI: Clustering-Guided Hyper-Connection Mamba for Hyperspectral Image Classification","created":"2026-03-03","cited_by_count":0,"area":"ssm"},{"id":"2603.16016","title":"FlatLands: Generative Floormap Completion From a Single Egocentric View","created":"2026-03-16","cited_by_count":0,"area":null},{"id":"2603.26756","title":"GradAttn: Replacing Fixed Residual Connections with Task-Modulated Attention Pathways","created":"2026-03-23","cited_by_count":0,"area":null},{"id":"2604.01210","title":"CliffSearch: Structured Agentic Co-Evolution over Theory and Code for Scientific Algorithm Discovery","created":"2026-04-01","cited_by_count":0,"area":"agents"},{"id":"2604.15069","title":"Beyond the Laplacian: Doubly Stochastic Matrices for Graph Neural Networks","created":"2026-04-16","cited_by_count":0,"area":null},{"id":"2604.19147","title":"Nexusformer: Nonlinear Attention Expansion for Stable and Inheritable Transformer Scaling","created":"2026-04-21","cited_by_count":0,"area":null},{"id":"2604.23705","title":"Can an MLP Absorb Its Own Skip Connection Exactly?","created":"2026-04-26","cited_by_count":0,"area":null},{"id":"2605.05330","title":"Graph Normalization: Fast Binarizing Dynamics for Differentiable MWIS","created":"2026-05-06","cited_by_count":0,"area":null},{"id":"2605.06501","title":"Cubit: Token Mixer with Kernel Ridge Regression","created":"2026-05-07","cited_by_count":0,"area":null},{"id":"2605.08300","title":"mHC-SSM: Manifold-Constrained Hyper-Connections for State Space Language Models with Stream-Specialized Adapters","created":"2026-05-08","cited_by_count":0,"area":null},{"id":"2605.11526","title":"Efficient and provably convergent end-to-end training of deep neural networks with linear constraints","created":"2026-05-12","cited_by_count":0,"area":null},{"id":"2605.15793","title":"AOT-POT: Adaptive Operator Transformation for Large-Scale PDE Pre-training","created":"2026-05-15","cited_by_count":0,"area":null},{"id":"2605.18848","title":"Exact Linear Attention","created":"2026-05-13","cited_by_count":0,"area":"moe"},{"id":"2605.23467","title":"S$^3$GNN: Efficient Global Mixing and Local Message Passing for Long-Range Graph Learning","created":"2026-05-22","cited_by_count":0,"area":null},{"id":"2606.05599","title":"Mitigating the Curse of Dimensionality in Uniform Convergence of Deep Neural Networks via Smooth Activations","created":"2026-06-04","cited_by_count":0,"area":null},{"id":"2606.07404","title":"Reversible Foundations: Training a 120B Sparse MoE through State-Preserving Scaling","created":"2026-06-05","cited_by_count":0,"area":"moe"},{"id":"2606.07419","title":"DisPOSE: Projected Polystochastic Diffusion for Self-Supervised Multi-View 3D Human Pose Estimation","created":"2026-06-05","cited_by_count":0,"area":null},{"id":"2606.19750","title":"Manifold Bandits: Bayesian Curriculum Learning over the Latent Geometry of Large Language Models","created":"2026-06-18","cited_by_count":0,"area":"lr"},{"id":"2607.04819","title":"Layer-Parallel Inference Reduces Encrypted Nonlinear Depth in Transformers","created":"2026-07-06","cited_by_count":0,"area":null},{"id":"2607.05148","title":"Fully Rotation-Equivariant Spectral-Spatial Learning for Multispectral Object Detection","created":"2026-07-06","cited_by_count":0,"area":null},{"id":"2607.28097","title":"From Expert Reduction to Behavioral Divergence: Tracing Numerical State through Sparse MoE Inference","created":"2026-07-30","cited_by_count":0,"area":"moe"},{"id":"2608.01283","title":"Riemannian Attention Mechanisms for Transformers: A Theoretical Framework and Architecture Design","created":"2026-08-02","cited_by_count":0,"area":null},{"id":"2608.08418","title":"Learning Deep Modality-Shared Self-Expressiveness for Image Clustering with Textual Information","created":"2026-08-09","cited_by_count":0,"area":"mllm"},{"id":"2608.09119","title":"Motif 3: Technical Report","created":"2026-08-10","cited_by_count":0,"area":"moe"},{"id":"2608.11760","title":"Tight Nonasymptotic Local Convergence of Sinkhorn-Knopp","created":"2026-08-12","cited_by_count":0,"area":null},{"id":"2608.18486","title":"WhiteMatter: All-to-All Cross-Layer Connections via KV Source Mixing","created":"2026-08-19","cited_by_count":0,"area":null},{"id":"2608.20069","title":"V-REX: Efficient Specialist VLM Training for Veterinary X-Rays","created":"2026-08-20","cited_by_count":0,"area":null},{"id":"2609.09306","title":"Gradland: On Phenomenal Experience, Differentiated Across Many Dimensions","created":"2026-09-08","cited_by_count":0,"area":null},{"id":"2609.09793","title":"How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE","created":"2026-09-09","cited_by_count":0,"area":"alignment"},{"id":"2609.33203","title":"Structured Residual Connectivity Matters for Diffusion Transformers","created":"2026-09-27","cited_by_count":0,"area":"diffusion"},{"id":"2609.33465","title":"GSM: Efficient Language Modeling with Shared Global State","created":"2026-09-27","cited_by_count":0,"area":null},{"id":"2609.33895","title":"Residual-Stream Burden Shapes Representation Learning in Diffusion Transformers","created":"2026-09-27","cited_by_count":0,"area":"diffusion"},{"id":"2609.37379","title":"Looped Transformers as Optimizers","created":"2026-09-29","cited_by_count":0,"area":null},{"id":"2609.39109","title":"T-Router: Learning Thalamic Routing for Reasoning with Parameter-Efficient Reinforcement Learning","created":"2026-09-30","cited_by_count":0,"area":null},{"id":"2610.01153","title":"Looping Beyond Twice: A Scalable Recipe for Looped Mixture-of-Experts","created":"2026-10-01","cited_by_count":0,"area":"moe"},{"id":"2610.06653","title":"The Birkhoff Geometry of Manifold-Constrained Hyper-Connections: Two Channels, Vertex Viscosity, and Sinkhorn as a Retraction","created":"2026-10-05","cited_by_count":0,"area":null},{"id":"2610.09179","title":"LayerRoPE: Dynamic Depth-wise Magnitude & Angular Superposition","created":"2026-10-06","cited_by_count":0,"area":null},{"id":"2610.13630","title":"Multi-Stream Attention Residuals","created":"2026-10-09","cited_by_count":0,"area":null},{"id":"2610.18407","title":"Multi-Stream Residual Computation in Diffusion Language Models","created":"2026-10-13","cited_by_count":0,"area":"diffusion"},{"id":"2610.18809","title":"Depth Memory","created":"2026-10-13","cited_by_count":0,"area":null},{"id":"2610.20350","title":"Lifted State Policy Optimization: Internalizing Self-Correction through Transition Geometry","created":"2026-10-14","cited_by_count":0,"area":"alignment"}],"next_cursor":null}