AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Meta-prediction Model for Distillation-Aware NAS on Unseen Datasets

Hayeon Lee, Sohyun An, Minseon Kim, Sung Ju Hwang

ICLR 2023 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 2 citations (OpenAlex)

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
encode the teacher and parameter-remapped student function instead of conditioning only on architecture or datasetDaSS · Unseen DaNAS tasks (CUB, Cars, DTD, Quick Draw)Spearman's rank correlation coefficient: 0.45 -> 0.64Improved
add teacher-accuracy-guided adaptation (ablation)DaSS · CUBSpearman's rank correlation coefficient: 0.69 -> 0.69No clear effect
add teacher-accuracy-guided adaptation (ablation)DaSS · Stanford CarsSpearman's rank correlation coefficient: 0.77 -> 0.88Improved
add teacher-accuracy-guided adaptation (ablation)DaSS · DTDSpearman's rank correlation coefficient: 0.50 -> 0.57Improved
add teacher-accuracy-guided adaptation (ablation)DaSS · Quick DrawSpearman's rank correlation coefficient: 0.62 -> 0.75Improved
add teacher-accuracy-guided adaptation (ablation)DaSS · Mean across CUB, Cars, DTD, and Quick DrawSpearman's rank correlation coefficient: 0.64 -> 0.72Improved
use DaSS to select the student architecture rather than the strongest meta-NAS baselineDaNAS · CUBaccuracy (%): 41.74 ± 0.34 -> 43.48 ± 0.11Improved
use DaSS to select the student architecture rather than the strongest meta-NAS baselineDaNAS · Stanford Carsaccuracy (%): 56.74 ± 0.36 -> 59.09 ± 0.06Improved
use DaSS to select the student architecture rather than the strongest meta-NAS baselineDaNAS · DTDaccuracy (%): 45.26 ± 0.78 -> 47.86 ± 0.40Improved
use DaSS to select the student architecture rather than the strongest meta-NAS baselineDaNAS · Quick Drawaccuracy (%): 46.80 ± 0.09 -> 47.27 ± 0.10Improved
use DaSS to select the student architecture rather than TANSDaNAS · CropDiseaseaccuracy (%): 99.72 ± 0.02 -> 99.71 ± 0.01No clear effect
use DaSS to select the student architecture rather than TANSDaNAS · EuroSATaccuracy (%): 97.33 ± 0.48 -> 97.95 ± 0.06Improved

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