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
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:17387-17410, 2026.

Abstract

Optimizing OS kernels for specific applications is vital for peak performance, yet existing LLM-based methods struggle with a semantic mismatch between generalized reasoning and low-level system behaviors. As a result, these static, open-loop approaches suffer from runtime blindness, configuration fragmentation, and search drift, ultimately failing to unlock the system’s performance potential. To address this, we propose QiMeng-PerceptOS, an autonomous framework that shifts the paradigm to perception-driven tuning. QiMeng-PerceptOS integrates: (1) a Perception Module that aligns raw telemetry into high-fidelity semantic fingerprints; (2) a Global Search Module utilizing a Bi-level Hierarchical Induction Tree (BHIT) for global navigation and efficient pruning; and (3) a Posterior Enhancement Module to suppress hallucinations via trajectory synthesis. Experiments across diverse workloads show that it achieves significant performance breakthroughs by optimizing kernel configurations, reaching 296.6% of default Redis throughput and surpassing SOTA baselines by 32.6% within only 15 iterations. By establishing a perception-driven closed-loop, QiMeng-PerceptOS provides new insights for fully automated, large-scale system optimization.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chen26ex, title = {{Q}i{M}eng-{P}ercept{OS}: Semantic-Aware Kernel Optimization for {OS}-Intensive Workloads via Hardware-Software Alignment}, author = {Chen, Huilai and Wen, Yuanbo and Li, Liangfeng and Peng, Shaohui and Zhu, Jingzhe and Bi, Jun and Zhang, Xuzhi and Guo, Qi and Li, Ling and Chen, Yunji}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {17387--17410}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/chen26ex/chen26ex.pdf}, url = {https://proceedings.mlr.press/v306/chen26ex.html}, abstract = {Optimizing OS kernels for specific applications is vital for peak performance, yet existing LLM-based methods struggle with a semantic mismatch between generalized reasoning and low-level system behaviors. As a result, these static, open-loop approaches suffer from runtime blindness, configuration fragmentation, and search drift, ultimately failing to unlock the system’s performance potential. To address this, we propose QiMeng-PerceptOS, an autonomous framework that shifts the paradigm to perception-driven tuning. QiMeng-PerceptOS integrates: (1) a Perception Module that aligns raw telemetry into high-fidelity semantic fingerprints; (2) a Global Search Module utilizing a Bi-level Hierarchical Induction Tree (BHIT) for global navigation and efficient pruning; and (3) a Posterior Enhancement Module to suppress hallucinations via trajectory synthesis. Experiments across diverse workloads show that it achieves significant performance breakthroughs by optimizing kernel configurations, reaching 296.6% of default Redis throughput and surpassing SOTA baselines by 32.6% within only 15 iterations. By establishing a perception-driven closed-loop, QiMeng-PerceptOS provides new insights for fully automated, large-scale system optimization.} }
Endnote
%0 Conference Paper %T QiMeng-PerceptOS: Semantic-Aware Kernel Optimization for OS-Intensive Workloads via Hardware-Software Alignment %A Huilai Chen %A Yuanbo Wen %A Liangfeng Li %A Shaohui Peng %A Jingzhe Zhu %A Jun Bi %A Xuzhi Zhang %A Qi Guo %A Ling Li %A Yunji Chen %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-chen26ex %I PMLR %P 17387--17410 %U https://proceedings.mlr.press/v306/chen26ex.html %V 306 %X Optimizing OS kernels for specific applications is vital for peak performance, yet existing LLM-based methods struggle with a semantic mismatch between generalized reasoning and low-level system behaviors. As a result, these static, open-loop approaches suffer from runtime blindness, configuration fragmentation, and search drift, ultimately failing to unlock the system’s performance potential. To address this, we propose QiMeng-PerceptOS, an autonomous framework that shifts the paradigm to perception-driven tuning. QiMeng-PerceptOS integrates: (1) a Perception Module that aligns raw telemetry into high-fidelity semantic fingerprints; (2) a Global Search Module utilizing a Bi-level Hierarchical Induction Tree (BHIT) for global navigation and efficient pruning; and (3) a Posterior Enhancement Module to suppress hallucinations via trajectory synthesis. Experiments across diverse workloads show that it achieves significant performance breakthroughs by optimizing kernel configurations, reaching 296.6% of default Redis throughput and surpassing SOTA baselines by 32.6% within only 15 iterations. By establishing a perception-driven closed-loop, QiMeng-PerceptOS provides new insights for fully automated, large-scale system optimization.
APA
Chen, H., Wen, Y., Li, L., Peng, S., Zhu, J., Bi, J., Zhang, X., Guo, Q., Li, L. & Chen, Y.. (2026). QiMeng-PerceptOS: Semantic-Aware Kernel Optimization for OS-Intensive Workloads via Hardware-Software Alignment. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:17387-17410 Available from https://proceedings.mlr.press/v306/chen26ex.html.

Related Material