AI Paper Atlas

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

ExVerus: Verus Proof Repair via Counterexample Reasoning

Jun Yang, Yuechun Sun, Yi Wu, Rodrigo Caridad, Yongwei Yuan, Jianan Yao, Shan Lu, Kexin Pei

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
Guide repair with source-level counterexamples (ExVerus) instead of AutoVerus handcrafted repair promptsVerus proof repair · VerusBenchsuccess rate (%): 24.7 -> 71.9 (DeepSeek-V3.1)Improved
Add counterexample-guided mutation and validation over direct fix from counterexamples (ablation)ExVerus · VerusBench (DeepSeek-V3.1)success rate (%): 64.4 -> 71.9Improved
Add counterexample-guided mutation and validation (ablation)ExVerus · ObfsBench (DeepSeek-V3.1)success rate (%): 65.4 -> 81.6Improved
Add mutation and validation (ablation)ExVerus · DafnyBench (o4-mini)success rate (%): 85.1 -> 95.5Improved
Add mutation and validation (ablation)ExVerus · DafnyBench (GPT-4o)success rate (%): 89.6 -> 88.1Hurt
Use 10 counterexamples instead of one (ablation)ExVerus · InvariantInjectBenchtasks proved (of 187): 100 -> 106Improved
Mutant blocks at least one validated counterexample (ablation)ExVerus mutant ranking · InvariantInjectBenchmutant verification rate: 38.55% -> 64.49%Improved
Use ExVerus under code obfuscation vs AutoVerusVerus proof repair · ObfsBenchsuccess rate (%): above 73% vs below 40%Improved
Use ExVerus instead of AutoVerus on costVerus proof repair · DeepSeek-V3.1, per taskcost (USD): 0.17 -> 0.04Improved
Use a stronger LLM (Sonnet-4.5) with ExVerus on easy benchmarksExVerus · DafnyBenchsuccess rate (%): 95.5 -> 95.5No clear effect
ExVerus on LCBench with weaker modelsExVerus · LCBench (DeepSeek-V3.1)success rate (%): 10.7 -> 10.7No clear effect

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