Research map / Architectures and efficiency / Attention and transformers
EcoFormer: Energy-Saving Attention with Linear Complexity
Jing Liu, Zizheng Pan, Haoyu He, Jianfei Cai, Bohan Zhuang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace MSA with EcoFormer binarized linear attention | PVTv2-B0 · ImageNet-1K attention only | on-chip energy (B pJ): 7.26 -> 0.50 | Improved |
| replace MSA with EcoFormer | PVTv2-B4 · ImageNet-1K attention only | on-chip energy (B pJ): 39.69 -> 10.52 | Improved |
| increase model size from PVTv2-B0 to B4 (ablation) | EcoFormer attention energy saving · ImageNet-1K | relative energy saving: -93% -> -74% | Hurt |
| replace MSA with EcoFormer | PVTv2-B0 · ImageNet-1K throughput | images/s: 850 -> 1379 | Improved |
| replace kernel-based linear attention (KLA) with EcoFormer | Twins-SVT-S · ImageNet-1K throughput | images/s: 489 -> 576 | Improved |
Later papers that compared with this method
1 later accepted paper reports Ecoformer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ShiftAddViT: Mixture of Multiplication Primitives Towards Efficient Vision Transformer (NeurIPS 2023 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Energy Transformer (NeurIPS 2023 · similarity 0.86)
- Token Statistics Transformer: Linear-Time Attention via Variational Rate Reduction (ICLR 2025 · similarity 0.84)
- BinaryAttention: One-Bit QK-Attention for Vision and Diffusion Transformers (CVPR 2026 · similarity 0.84)
- MemoryFormer : Minimize Transformer Computation by Removing Fully-Connected Layers (NeurIPS 2024 · similarity 0.83)
- Fast attention mechanisms: a tale of parallelism (NeurIPS 2025 · similarity 0.83)
- Learning Linear Attention in Polynomial Time (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
