AI Paper Atlas

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

ParEVO: Synthesizing Code for Irregular Data: High-Performance Parallelism through Agentic Evolution

Liu Yang, Zeyu Nie, Andrew Liu, Ruomu Zou, Deniz Altınbüken, Amir Yazdanbakhsh, Quanquan C. Liu

ICML 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics · 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
add evolutionary coding agent with 10 iterations (ablation)Gemini-3-Pro · DMOJ training problemsspeedup: 1.00 -> 1.498Improved
add evolutionary coding agent with 30 iterations (ablation)Gemini-3-Pro · DMOJ training problemsspeedup: 1.00 -> 2.218Improved
fine-tune on Parlay-Instruct and DMOJGemini-2.5-Pro · ParEval graph problemsPass@1: 0.42 -> 0.76Improved
fine-tune on Parlay-Instruct and DMOJGemini-2.5-Pro · ParEval graph problemsSpeedup@1: 21.76 -> 13.67Hurt
fine-tune on Parlay-Instruct and DMOJGemini-2.5-Pro · ParEval graph problemsBuild@1: 0.62 -> 0.97Improved
fine-tune on Parlay-Instruct and DMOJGemini-2.5-Pro · ParEval C++ ParlayLibSpeedup@1 (AM): 85.34 -> 107.43Improved
fine-tune on Parlay-Instruct and DMOJGemini-2.5-Pro · ParEval C++ ParlayLibPass@1: 0.54 -> 0.58Improved
fine-tune on ParlayLib syntax and slow-fast pairsDeepSeek-6.7B-Base · ParEval C++ ParlayLibSpeedup@1 (AM): 0.73 -> 126.16Improved
fine-tune on ParlayLib syntax and slow-fast pairsDeepSeek-6.7B-Base · ParEval C++ ParlayLibPass@1: 0.11 -> 0.26Improved
fine-tune with SFT and DPO for RustQwen3-Coder-30B-Instruct · ParEval Rust RayonSpeedup@1 (AM): 4.57 -> 4.62No clear effect
fine-tune on Parlay dataGemini-2.5-Pro · ParEval convex hullPass@1: 0.45 -> 0.00Hurt
fine-tune on Parlay dataGemini-2.5-Pro · ParEval closest pair 2DSpeedup@1: 74.48 -> 188.03Improved

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