AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Conditional Adapters: Parameter-efficient Transfer Learning with Fast Inference

Tao Lei, Junwen Bai, Siddhartha Brahma, Joshua Ainslie, Kenton Lee, Yanqi Zhou, Nan Du, Vincent Y Zhao, Yuexin Wu, Bo Li, Yu Zhang, Ming-Wei Chang

NeurIPS 2023 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging

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 changedWhereMeasuredEffect
Apply CoDA with k-to-all attention and reduction r=3 instead of dense Parallel AdapterT5 v1.1 Base, Large, and XL · MNLI, RTE, BoolQaccuracy: 86.6 vs 87.1; 72.6 vs 71.5; 76.6 vs 77.9 (Base); +0.2 Δ AvgMixed
Increase reduction factor from r=3 to r=5 (ablation)CoDA k-to-all · MNLI, RTE, BoolQ across T5 v1.1 sizesaverage accuracy: +0.2 -> −0.6 Δ AvgHurt
Use k-to-all attention rather than k-to-k attention (ablation)CoDA · MNLI, RTE, BoolQ across T5 v1.1 sizesaverage accuracy: −0.3 -> +0.2 Δ Avg at r=3; −2.1 -> −0.6 Δ Avg at r=5Improved
Use soft top-k routing instead of independent sigmoid gates (ablation)CoDA · MNLI, RTE, BoolQ across T5 v1.1 sizesaverage accuracy: −2.0 at r=3; −1.6 at r=5Improved
Use learned soft top-k routing instead of selecting the first k tokens (ablation)CoDA · MNLI, RTE, BoolQ across T5 v1.1 sizesaverage accuracy: −4.4 at r=3; −12.9 at r=5Improved
Use CoDA with reduction r=3 rather than dense Parallel AdapterT5 v1.1 models · MNLI, RTE, BoolQTransformer computation: 3-5x computation reductionImproved
Pretrain CoDA for more steps (ablation)CoDA initialized from pretrained dense models · Downstream classification tasksaccuracyImproved
Apply conditional token computationCoDA versus Parallel Adapter · NLP tasksinference speed: 2x to 8xImproved
Combine conditional computation with LoRA at reduction r=3LoRA · Six language tasksaverage score: 84.0 vs 84.0No clear effect
Combine conditional computation with LoRA at reduction r=5LoRA · Six language tasksaverage score: 84.0 vs 83.7Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy