Research map / Architectures and efficiency / Pruning and efficient tuning
Compacter: Efficient Low-Rank Hypercomplex Adapter Layers
Rabeeh Karimi mahabadi, James Henderson, Sebastian Ruder
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 |
|---|---|---|---|
| Use shared hypercomplex weights and layer-specific rank-one factors in adapters | T5-Base fine-tuning on GLUE · GLUE | average score: 86.50 -> 86.62 | Improved |
| Use shared hypercomplex weights and layer-specific rank-one factors in adapters | T5-Base fine-tuning on SuperGLUE · SuperGLUE | average score: 70.06 -> 71.57 | Improved |
| Remove the Compacter layer after self-attention (ablation) | Compacter on GLUE · GLUE | average score: 86.62 -> 86.47 | No clear effect |
| Remove the Compacter layer after self-attention (ablation) | Compacter on SuperGLUE · SuperGLUE | average score: 71.57 -> 71.82 | Improved |
| Set n=8 instead of n=4 (ablation) | Compacter on GLUE · GLUE | average score: 86.62 -> 86.34 | Hurt |
| Set n=12 instead of n=4 (ablation) | Compacter on GLUE · GLUE | average score: 86.62 -> 86.17 | Hurt |
| Set n=8 instead of n=4 (ablation) | Compacter++ on GLUE · GLUE | average score: 86.47 -> 85.95 | Hurt |
| Set n=12 instead of n=4 (ablation) | Compacter++ on GLUE · GLUE | average score: 86.47 -> 86.34 | Hurt |
| Set n=8 instead of n=12 (ablation) | Compacter on SuperGLUE · SuperGLUE | average score: 71.57 -> 71.19 | Hurt |
| Set n=4 instead of n=12 (ablation) | Compacter on SuperGLUE · SuperGLUE | average score: 71.57 -> 70.18 | Hurt |
| Set n=8 instead of n=12 (ablation) | Compacter++ on SuperGLUE · SuperGLUE | average score: 71.82 -> 68.69 | Hurt |
| Set n=4 instead of n=12 (ablation) | Compacter++ on SuperGLUE · SuperGLUE | average score: 71.82 -> 70.53 | Hurt |
Closest papers by meaning
- SoLA: Leveraging Soft Activation Sparsity and Low-Rank Decomposition for Large Language Model Compression (AAAI 2025 · similarity 0.89)
- SPP: Sparsity-Preserved Parameter-Efficient Fine-Tuning for Large Language Models (ICML 2024 · similarity 0.89)
- Benchmarking Optimizers for Large Language Model Pretraining (NeurIPS 2026 · similarity 0.89)
- Toward Adaptive Large Language Models Structured Pruning via Hybrid-grained Weight Importance Assessment (AAAI 2025 · similarity 0.88)
- Enhancing Large Language Model Performance with Gradient-Based Parameter Selection (AAAI 2025 · similarity 0.88)
- LensLLM: Unveiling Fine-Tuning Dynamics for LLM Selection (ICML 2025 · similarity 0.88)
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.
