AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

HyperTree Proof Search for Neural Theorem Proving

Guillaume Lample, Timothee Lacroix, Marie-anne Lachaux, Aurelien Rodriguez, Amaury Hayat, Thibaut Lavril, Gabriel Ebner, Xavier Martinet

NeurIPS 2022 · Architectures and efficiency › Neural architecture search · approach: AutoML pipeline search · 25 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 online training on proof-search data to supervised modelEvariste on Metamath · Metamath test pass@32pass rate %: 61.2% -> 72.4%Improved
add online training on unproved statementsEvariste on Equations Identities · Equations Identitiescumulative pass rate %: never exceeds 36% -> 91.3%Improved
train policy on minimal proofs instead of all proofs of solved nodes (ablation)Evariste · Equations Identitiescumulative pass rate %: 40.6 -> 78.1Improved
train policy on root minimal proof instead of all nodes (ablation)Evariste · Metamath validpass@8 %: 51.6 -> 68.6Improved
restrict to minimal proofs on Metamath test (ablation)Evariste · Metamath testpass@8 %: 57.2 -> 58.8Improved
add critic model (ablation)Evariste · Equations Identitiescumulative pass rate %: 65.6 -> 78.1Improved
add critic model (ablation)Evariste · Metamath validpass@8 %: 64.8 -> 68.6Improved
soft critic targets (HTPS action values) instead of hard 0/1 targets (ablation)Evariste · Equations Identitiescumulative pass rate %: 63.1 -> 78.1Improved
sample search hyper-parameters per attempt instead of fixed (ablation)Evariste · Equations Identitiescumulative pass rate %: 73.8 -> 78.1Improved
sample search hyper-parameters per attempt instead of fixed (ablation)Evariste · Metamath validpass@8 %: 69.8 -> 68.6 valid; 56.2 -> 57.4 testMixed
add synthetic Equations co-training task (ablation)Evariste on Lean · miniF2F-valid cumulativecumulative pass rate %: 54.9% -> 58.6%Improved
refresh model on provers more frequently (online vs expert iteration) (ablation)Evariste · Equations/Metamathcumulative pass rateImproved

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