AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Graph-based, Self-Supervised Program Repair from Diagnostic Feedback

Michihiro Yasunaga, Percy Liang

ICML 2020 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Self-supervision in supervised pipelines · 63 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
Add the program-feedback graph to the sequence model (ablation)DrRepair · DeepFix raw testfull repair rate: 62.5 -> 66.4Improved
Add self-supervised pre-training (ablation)Base + graph · DeepFix raw testfull repair rate: 66.4 -> 68.2Improved
Remove compiler messages (ablation)Base repair model · DeepFix raw testfull repair rate: 62.5 -> 34.0Hurt
Add the program-feedback graph (ablation)SPoC repair model · SPoC devsingle repair accuracy: 48.6 -> 53.0Improved
Add self-supervised pre-training (ablation)SPoC Base + graph · SPoC devsingle repair accuracy: 53.0 -> 56.2Improved
Provide pseudocode as an additional input (ablation)DrRepair · SPoC devsingle repair accuracy: 56.2 -> 68.0Improved
Use DrPerturb instead of Gupta+17 corruption (ablation)Base repair model · DeepFix raw testfull repair rate: 50.5 -> 62.5Improved
Use DrPerturb instead of random token-drop corruption (ablation)Base repair model · DeepFix raw testfull repair rate: 49.4 -> 62.5Improved
Replace the program-feedback graph with code-only edges (ablation)Base + graph · DeepFix raw testfull repair rate: 66.4 -> 64.8Hurt
Replace the program-feedback graph with code-feedback edges only (ablation)Base + graph · DeepFix raw testfull repair rate: 66.4 -> 64.9Hurt
Replace the program-feedback graph with self-attention (ablation)Base repair model · DeepFix raw testfull repair rate: 66.4 -> 66.0Hurt
Add DrRepair to best-first program synthesis searchSPoC synthesis · SPoC TestP, B=100synthesis success rate: 34.2 -> 37.5Improved

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