AI Paper Atlas

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

LogiGAN: Learning Logical Reasoning via Adversarial Pre-training

Xinyu Pi, Wanjun Zhong, Yan Gao, Nan Duan, Jian-Guang Lou

NeurIPS 2022 · Large language models › Language, safety and interpretability · approach: Faithfulness of LLM explanations · 7 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
pre-train T5 base with LogiGAN (ss+es)T5 base · 12 downstream tasks (classification avg)accuracy: 49.52 -> 54.62Improved
pre-train T5 base with LogiGAN (ss+es)T5 base · generation tasks avgscore: 36.65 -> 37.77Improved
pre-train T5 large with LogiGANT5 large · classification tasks avgaccuracy: 63.56 -> 66.47Improved
pre-train T5 large with LogiGANT5 large · generation tasks avgscore: 38.87 -> 40.28Improved
pre-train T5 large with LogiGANT5 large · ReCloraccuracy: 50.40 -> 54.80Improved
add Elastic Search retrieved pseudo-statements to self-samplingLogiGAN base · 12 tasks classification avgaccuracy: 54.06 -> 54.62Improved
add Elastic Search retrieved pseudo-statements to self-samplingLogiGAN base · LogiQAaccuracy: 34.72 -> 37.02Improved
add Elastic Search retrieved pseudo-statements to self-samplingLogiGAN base · ReCloraccuracy: 40.20 -> 40.00No clear effect
replace logical statements with random masked sentences (I) (ablation)LogiGAN base · 7-dataset averageaccuracy: 49.80 -> 50.10No clear effect
MLE-only logic pre-training without adversarial training (II) (ablation)T5 base · 7-dataset averageaccuracy: 49.80 -> 52.71Improved
add adversarial Generator-Verifier training on top of MLE-only (full LogiGAN vs II) (ablation)MLE logic pre-train · 7-dataset averageaccuracy: 52.71 -> 54.59Improved
iterative multi-task generation plus verification in one model (III) vs MLE-only (II) (ablation)MLE logic pre-train · 7-dataset averageaccuracy: 52.71 -> 52.08Hurt

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