AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Learning to Scaffold: Optimizing Model Explanations for Teaching

Patrick Fernandes, Marcos Vinicius Treviso, Danish Pruthi, Andre Martins, Graham Neubig

NeurIPS 2022 · Large language models › Language, safety and interpretability · approach: Data and evaluation methodology · 10 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
learn head coefficients with SMaT instead of static mean over all headsattention explainer · IMDB, 500 samplessimulability accuracy (%): 83.00 -> 91.48Improved
learn head coefficients with SMaT instead of static all-layer attentionattention explainer · IMDB, 2,000 samplessimulability accuracy (%): 90.08 -> 92.84Improved
learn head coefficients with SMaT instead of static attentionattention explainer · CIFAR-100, 9,000 samplessimulability accuracy (%): 85.92 -> 85.96Improved
use SMaT-learned attention instead of Gradient L2 at 8,400 samples (median)attention explainer · MLQE-PEPearson correlation: .8638 vs .8561Mixed
add gradient x input or integrated gradients explanations to student trainingstudent · IMDB, 1,000 samplessimulability accuracy (%): 83.44 -> 81.15Hurt
use only last-layer heads instead of all heads (static attention) (ablation)attention explainer · IMDB, 500 samplessimulability accuracy (%): 83.00 -> 80.91Hurt
use Softmax instead of Sparsemax for head coefficient normalization (ablation)SMaT · MLQE-PE 4,200Pearson correlation: .8630 -> .7439Hurt
use 1.5-Entmax instead of Sparsemax (ablation)SMaT · MLQE-PE 4,200Pearson correlation: .8630 -> .7163Hurt
remove head coefficient normalization (ablation)SMaT · MLQE-PE 4,200Pearson correlation: .8630 -> .8002Hurt
train explainer with SMaT instead of static all-layer attentionattention explainer · MovieReviews plausibilityAUC: 0.68 -> 0.73Improved
train explainer with SMaT instead of static all-layer attentionattention explainer · CIFAR-100 user studyTrueSkill: 0.7±.67 -> 4.3±.70Improved

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