TPAMI 2025 papers on Architectures and efficiency
44 of the 932 papers published in TPAMI 2025 (4.7%) are on Architectures and efficiency, compared with 6.2% at TPAMI 2024. They fall into 8 topics.
See TPAMI 2025 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Attention and transformers · 14 papers
- Fully-Connected Transformer for Multi-Source Image Fusion (53 citations)
- S2-Transformer for Mask-Aware Hyperspectral Image Reconstruction (34 citations)
- Learning an Adaptive Sparse Transformer for Efficient Image Restoration (23 citations)
- Efficient Visual Transformer by Learnable Token Merging (15 citations)
- Complexity Control Facilitates Reasoning-Based Compositional Generalization in Transformers (14 citations)
- DHVT: Dynamic Hybrid Vision Transformer for Small Dataset Recognition (14 citations)
- EMOv2: Pushing 5M Vision Model Frontier (10 citations)
- The CUR Decomposition of Self-Attention Matrices in Vision Transformers (9 citations)
- Taylor-Series-Expansion-Based Vision Transformer Models (5 citations)
- HPformer: Low-Parameter Transformer With Temporal Dependency Hierarchical Propagation for Health Informatics
- Bayesian Window Transformer for Image Restoration
- CAIT: Triple-Win C ompression Toward High A ccuracy, Fast I nference, and Favorable T ransferability for ViTs
and 2 more in the interactive map.
Pruning and efficient tuning · 9 papers
- RingMo-Aerial: An Aerial Remote Sensing Foundation Model With Affine Transformation Contrastive Learning (17 citations)
- Zero-Shot Sparse Mixture of Low-Rank Experts Construction From Pre-Trained Foundation Models (14 citations)
- PRANCE: Joint Token-Optimization and Structural Channel-Pruning for Adaptive ViT Inference (13 citations)
- Stimulative Training++: Go Beyond the Performance Limits of Residual Networks (9 citations)
- Efficient Distortion-Minimized Layerwise Pruning (3 citations)
- DTL: Parameter- and Memory-Efficient Disentangled Vision Learning (1 citation)
- Toward Unified Expertise: Learning a Single Vision Model From Diverse Perception
- Pruning at Initialization – A Sketching Perspective
- Spatial Re-Parameterization for N:M Sparsity
Quantization · 6 papers
- Towards Accurate Post-Training Quantization of Vision Transformers via Error Reduction (27 citations)
- I&S-ViT: An Inclusive & Stable Method for Post-Training ViTs Quantization (25 citations)
- LR Quant+ : A Unified and Learnable Framework to Post-Training Quantization for Transformer-Based Large Foundation Models (2 citations)
- Latent Weight Quantization for Integerized Training of Deep Neural Networks
- NUPES: Non-Uniform Post-Training Quantization via Power Exponent Search
- Pushing the Limit of Post-Training Quantization
CNN design · 5 papers
- Scaling Up Your Kernels: Large Kernel Design in ConvNets Toward Universal Representations (38 citations)
- MOL: Joint Estimation of Micro-Expression, Optical Flow, and Landmark via Transformer-Graph-Style Convolution (28 citations)
- Deep Lookup Network (4 citations)
- ACLI: A CNN Pruning Framework Leveraging Adjacent Convolutional Layer Interdependence and $\gamma$γ-Weakly Submodularity (3 citations)
- Interpretable Rotation-Equivariant Multiary-Valued Network for Attribute Obfuscation
Neural architecture search · 5 papers
- BossNAS Family: Block-Wisely Self-Supervised Neural Architecture Search (10 citations)
- $\beta $-DARTS++: Bi-Level Regularization for Proxy-Robust Differentiable Architecture Search (9 citations)
- Accelerating Zero-Shot NAS With Feature Map-Based Proxy and Operation Scoring Function (1 citation)
- EA-HAS-Bench and Language-Enhanced Shrinkage Search for Energy-Aware NAS (1 citation)
- Good Performance Estimation Strategies are All You Need in Neural Architecture Search
RNNs and normalization · 3 papers
- One Neuron Saved is One Neuron Earned: On Parametric Efficiency of Quadratic Networks (18 citations)
- On the Upper Bounds of Number of Linear Regions and Generalization Error of Deep Convolutional Neural Networks (3 citations)
- On the Universal Approximation Properties of Deep Neural Networks Using MAM Neurons
