AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Divide and Abstract: Autoformalization via Decomposition and Abstraction Learning

Marcus J. Min, Yeqi Gao, Wilson Sy, Zhaoyu Li, Xujie Si, Osbert Bastani

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data

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 learned abstractions and decomposition (full DNA) to the baselinenon-reasoning models average · LeanEuclidPluspass@1 accuracy: 18.7 -> 40.6Improved
add learned abstractions and decomposition (full DNA) to the baselinereasoning models average · LeanEuclidPluspass@1 accuracy: 31.5 -> 51.1Improved
add decomposition (Divide) to the baseline (ablation)non-reasoning models · LeanEuclidPluspass@1 accuracy: 18.7 -> 28.9Improved
add learned abstractions (Abstract) to the baseline (ablation)non-reasoning models · LeanEuclidPluspass@1 accuracy: 18.7 -> 20.9Improved
add decomposition on top of learned abstractions (Abstract -> DNA) (ablation)non-reasoning models · LeanEuclidPluspass@1 accuracy: 20.9 -> 40.6Improved
add learned abstractions on top of decomposition (Divide -> DNA) (ablation)reasoning models · LeanEuclidPluspass@1 accuracy: 36.5 -> 51.1Improved
add decomposition to Qwen3-14BQwen3 14B · LeanEuclidPluspass@1 accuracy: 1.0 -> 9.6Improved
add DNA to baseline on ProofNet-HardGPT-5 mini · ProofNet-Hardpass@1 accuracy: 0.0 -> 22.1 (DNOracleA)Improved
add decomposition prompts to specialized autoformalizers (Abstract -> DNA) (ablation)Kimina Autoformalizer 7B · ProofNet-Hardpass@1 accuracy: 8.4 -> 1.1Hurt
add decomposition to abstractions on harder problems (Abstract -> DNA) (ablation)gpt-5-mini · LeanEuclidPlus (40 complex)pass@1 accuracy: 47.5 -> 57.5Improved
add 1-shot example (vs 0-shot) to DNA (ablation)non-reasoning models average · LeanEuclidPluspass@1 accuracy: 19.3 -> 40.6Improved
add 1-shot example (vs 0-shot) to DNA (ablation)reasoning models average · LeanEuclidPluspass@1 accuracy: 26.0 -> 51.1Improved

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