Research map / Architectures and efficiency / Attention and transformers
Paramixer: Parameterizing Mixing Links in Sparse Factors Works Better Than Dot-Product Self-Attention
Tong Yu, Ruslan Khalitov, Lei Cheng, Zhirong Yang
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 |
|---|---|---|---|
| Parameterize mixing links in sparse full-rank factors instead of using dot-product attention | ParamixerNet · Long Range Arena Text | accuracy: 68.4% -> 83.32% | Improved |
| Use CHORD rather than CDIL sparse-link protocol (ablation) | Paramixer · Long Range Arena Pathfinder | accuracy: 67.13±0.42 -> 80.49±0.13 | Improved |
| Use CDIL rather than CHORD sparse-link protocol (ablation) | Paramixer · Long Range Arena Image | accuracy: 45.01±0.21 -> 46.58±0.05 | Improved |
| Use CDIL rather than CHORD sparse-link protocol (ablation) | Paramixer · Long Range Arena Text | accuracy: 83.12±0.33 -> 83.32±0.19 | No clear effect |
| Use CHORD rather than CDIL sparse-link protocol (ablation) | Paramixer · Long Range Arena ListOps | accuracy: 37.78±0.28 -> 39.57±0.32 | Improved |
| Parameterize mixing links from the original input rather than the preceding block's output (ablation) | ParamixerNet · Paramixer neural-network experiments | training performance | Improved |
| Use longer context (32k rather than 16k tokens) (ablation) | Paramixer · Long Document | accuracy: 83.89 -> 84.55 | Improved |
| Parameterize sparse-factor mixing links instead of dot-product attention | Paramixer · HFDNA | ROC AUC: 94.32 -> 100.00 | Improved |
| Parameterize sparse-factor mixing links instead of dot-product attention | Paramixer · MTcDNA | ROC AUC: 79.38 -> 84.86 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Nyströmformer — Nyströmformer: A Nyström-based Algorithm for Approximating Self-Attention (AAAI 2021) · better on 2 of 2 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
- Delta Attention: Fast and Accurate Sparse Attention Inference by Delta Correction (NeurIPS 2025 · similarity 0.83)
- TdAttenMix: Top-Down Attention Guided Mixup (AAAI 2025 · similarity 0.83)
- Two Sparse Matrices are Better than One: Sparsifying Neural Networks with Double Sparse Factorization (ICLR 2025 · similarity 0.83)
- Group and Shuffle: Efficient Structured Orthogonal Parametrization (NeurIPS 2024 · similarity 0.83)
- SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference (ICML 2025 · similarity 0.83)
- Bayes optimal learning of attention-indexed models (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.
