AI Paper Atlas

Research map / Large language models / Agents, code and math

Compile to Compress: Boosting Formal Theorem Provers by Compiler Outputs

Guchan Li, Rui Tian, Hongning Wang

ICML 2026 · Large language models › Agents, code and math · approach: Retrieval-augmented formal theorem proving · 0 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
fine-tune on refinement data with expert iteration plus random tree searchGoedel-Prover-V2-32B · PutnamBench (budget 64)problems solved: 32 -> 63Improved
fine-tune on refinement data with expert iteration plus random tree searchKimina-Distill-8B · PutnamBench (budget 64)problems solved: 10 -> 17Improved
replace refinement training with direct-solution SFT on more data (ablation)Kimina-Distill-8B · MiniF2F-testaccuracy [%]: 77.46 -> 75.41Hurt
replace refinement training with direct-solution SFT on more data (ablation)Goedel-Prover-V2-32B · MathOlympiadBenchaccuracy [%]: 14.72 -> 12.50Hurt
add value-guided tree search over random tree search (ablation)Goedel-Prover-V2-32B · MiniF2F-testaccuracy [%]: 84.43 -> 86.89Improved
add value-guided tree search over random tree search (ablation)Kimina-Distill-8B · ProofNetaccuracy [%]: 15.63 -> 15.36Hurt
add value-guided search on hard problems (ablation)Goedel-Prover-V2-32B · MiniF2F-test hardaccuracy [%]: 18.42 -> 26.32Improved
use DFS refinement instead of BFS sampling (ablation)Kimina-Distill-8B · MiniF2F-test hardaccuracy [%]: 10.91 -> 20.00Improved
use refinement instead of direct sampling on easy problemsprovers · MiniF2F-testaccuracy [%]No clear effect
scale test budget from 64 to 256Goedel-Expert (value) · PutnamBenchproblems solved: 63 -> 110Improved
use Markovian refinement with Goedel-ExpertGoedel-Prover-V2-32B · PutnamBench total output tokenstokens: 924,911 K -> 476,133 KImproved

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