Research map / Architectures and efficiency / Mixture of experts
Alternating Updates for Efficient Transformers
Cenk Baykal, Dylan J Cutler, Nishanth Dikkala, Nikhil Ghosh, Rina Panigrahy, Xin Wang
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 |
|---|---|---|---|
| Add AltUp (K=2) | T5-Base · GLUE | score: 84.25 -> 85.32 | Improved |
| Add AltUp (K=2) | T5-Base · SuperGLUE | score: 73.56 -> 75.80 | Improved |
| Increase K from 2 to 4 (ablation) | AltUp on T5-Small · GLUE | score: 76.82 -> 76.40 | Hurt |
| Increase K from 2 to 4 (ablation) | AltUp on T5-Large · GLUE | score: 88.20 -> 88.42 | Improved |
| Increase K from 2 to 4 (ablation) | AltUp on T5-Small · pretrain accuracy | accuracy: 61.86 -> 62.00 | Improved |
| Use Sequence-AltUp instead of Stride-and-Skip | T5-Base · SuperGLUE finetune | score: 65.98 -> 66.94 | Improved |
| Use average pooling for sequence reduction | T5-Base · SuperGLUE finetune | score: 73.56 -> 57.85 | Hurt |
| Use AltUp instead of dense 2x width scaling | T5-Base · pretrain accuracy and train speed | accuracy: 66.45 -> 65.78 | Mixed |
| Combine AltUp with MoE | T5-Base · 100k steps pretrain | accuracy: 63.97 -> 64.13 | Improved |
| Add Recycled-AltUp (K=2) | T5-Base · GLUE | score: 84.25 -> 85.60 | Improved |
| Use Recycled-AltUp instead of AltUp | T5-Small · pretrain accuracy | accuracy: 61.86 -> 61.33 | Hurt |
| Replace Sum with Prediction-Compute-Correct (AltUp) (ablation) | T5-Large · GLUE | score: 86.18 -> 88.20 | Improved |
Closest papers by meaning
- HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization (NeurIPS 2025 · similarity 0.87)
- DenseFormer: Enhancing Information Flow in Transformers via Depth Weighted Averaging (NeurIPS 2024 · similarity 0.86)
- Trainable Transformer in Transformer (ICML 2024 · similarity 0.86)
- StarTrail: Concentric Ring Sequence Parallelism for Efficient Near-Infinite-Context Transformer Model Training (NeurIPS 2025 · similarity 0.86)
- SwitchHead: Accelerating Transformers with Mixture-of-Experts Attention (NeurIPS 2024 · similarity 0.86)
- A Multi-Level Framework for Accelerating Training Transformer Models (ICLR 2024 · similarity 0.86)
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.
