AI Paper Atlas

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

AlgoVeri: An Aligned Benchmark for Verified Code Generation on Classical Algorithms

Haoyu Zhao, Ziran Yang, Jiawei Li, Mike He, Zenan Li, Chi Jin, Venugopal Veeravalli, Aarti Gupta, Sanjeev Arora

ICML 2026 · Large language models › Agents, code and math · approach: Software engineering agents · 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
increase parallel sampling from 1 to 10 chains of 15 repair rounds (ablation)GPT-OSS-120B · AlgoVeri Dafnyfull mark pass rate (%): 13.51 -> 28.57Improved
increase parallel sampling from 1 to 10 chains of 15 repair rounds (ablation)Qwen3-235B-A22B-Instruct · AlgoVeri Dafnyfull mark pass rate (%): 18.31 -> 29.87Improved
increase parallel sampling from 1 to 10 chains of 15 repair rounds (ablation)Qwen3-Next-80B-A3B-Thinking · AlgoVeri Dafnyfull mark pass rate (%): 17.27 -> 33.77Improved
increase parallel sampling from 1 to 10 chains of 15 repair rounds (ablation)Devstral-2-123B-Instruct · AlgoVeri Dafnyfull mark pass rate (%): 6.10 -> 18.18Improved
increase parallel sampling from 1 to 10 chains of 15 repair rounds (ablation)GPT-OSS-120B · AlgoVeri Leanfull mark pass rate (%): 7.01 -> 14.29Improved
increase parallel sampling from 1 to 10 chains of 15 repair rounds (ablation)Qwen3-235B-A22B-Instruct · AlgoVeri Verusfull mark pass rate (%): 9.09 -> 12.99Improved
add semantic filtering by an LLM judge to compiler-verified resultsGemini-3 Flash · AlgoVeri Dafnypass rate (%): 55.84 -> 40.26Hurt
add semantic filtering by an LLM judge to compiler-verified resultsGPT-5 mini · AlgoVeri Dafnypass rate (%): 41.56 -> 30.47Hurt
add semantic filtering by an LLM judge to compiler-verified resultsGPT-5.3 Codex · AlgoVeri Leanpass rate (%): 23.38 -> 11.69Hurt
add iterative repair rounds with compiler feedback (ablation)Gemini-3 Flash · AlgoVeri Dafnyaverage pass rate: nearly triples from Round 0 to 15Improved
add iterative repair rounds with compiler feedback (ablation)GPT-OSS-120B · AlgoVeriaverage pass rate: saturates by Round 3No clear effect
replace pure sampling with repair chains at equal total budget (N/15 chains of 15 rounds) (ablation)GPT-OSS-120B · AlgoVeripass rateHurt

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