AI Paper Atlas

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

QiMeng-PerceptOS: Semantic-Aware Kernel Optimization for OS-Intensive Workloads via Hardware-Software Alignment

Huilai Chen, Yuanbo Wen, Liangfeng Li, Shaohui Peng, Jingzhe Zhu, Jun Bi, Xuzhi Zhang, Qi Guo, Ling Li, Yunji Chen

ICML 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics

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 hardware fingerprint to the perception module (ablation)QiMeng-PerceptOS · Apache (Ubuntu workstation)performance relative to default: 114.34% -> 128.90%Improved
add software fingerprint (hot call stacks) (ablation)QiMeng-PerceptOS · Apache (Ubuntu workstation)performance relative to default: 116.41% -> 128.90%Improved
add hierarchical induction summaries (ablation)QiMeng-PerceptOS · Apache (Ubuntu workstation)performance relative to default: 121.70% -> 128.90%Improved
add elite configuration memory and closed-loop feedback (ablation)QiMeng-PerceptOS · Apache (Ubuntu workstation)performance relative to default: 116.70% -> 128.90%Improved
replace 2 global search rounds with 2 posterior enhancement rounds (ablation)QiMeng-PerceptOS · Redis (Ubuntu PC)performance relative to default: 260.95% -> 278.53%Improved
replace 2 global search rounds with 2 posterior enhancement rounds (ablation)QiMeng-PerceptOS · Apache and RAG (Ubuntu PC)performance relative to default: 295.89% -> 295.89%No clear effect
replace 2 global search rounds with 2 posterior enhancement rounds (ablation)QiMeng-PerceptOS · PostgreSQL (Ubuntu PC)performance relative to default: 149.88% -> 154.50%Improved
add LLM-based semantic discriminator window (k=0 to k=50 with v=0) (ablation)perception module · Apache (Ubuntu workstation)performance relative to default: 116.4% -> 124.6%Improved
add fixed top-v stacks (v=0 to v=20, k=50) (ablation)perception module · Apache (Ubuntu workstation)performance relative to default: 124.6% -> 128.9%Improved
enlarge perception window from v=20,k=50 to v=40,k=80 (ablation)perception module · Apache (Ubuntu workstation)performance relative to default: 128.9% -> 127.5%No clear effect
use perception-driven BHIT search instead of static AutoOS promptskernel tuning with GPT-4o-mini · RAG (Fedora 42 PC Intel)performance relative to default: 102.2% -> 133.8%Improved
use QiMeng-PerceptOS instead of AutoOSkernel tuning with DeepSeek-V3.2 · RAG (Fedora 42 PC Intel)improvement vs Vanilla: Vanilla 136.5% vs QiMeng-PerceptOS 133.7%Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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