# EngineAgent: Agentizing On-Device LLM Inference Engines with Inference Stack Hierarchy Awareness and Hardware Abilities

**Authors:** Zhengxiang Huang, Yujie Sun, Shengheng Chen, Zhaode Wang, Chaoyue Niu, Zeyu Zhao, Chengfei Lv, Fan Wu, Guihai Chen  
**Published:** 10 Nov 2026  
**Links:** [arXiv](https://arxiv.org/abs/2611.14356) · [PDF](https://arxiv.org/pdf/2611.14356) · [This page](https://omen-bit.github.io/ai-paper-daily/paper/2611.14356.html)

## Key numbers

- **75%**: EngineAgent achieves 75% end-to-end success with either of two underlying LLMs
- **33.3%**: versus 33.3% for the strongest coding-agent baseline
- **12**: Across 12 tasks spanning three mobile LLMs and four engine--backend pairs on CPUs

## Abstract

Rapidly evolving LLM architectures and heterogeneous mobile hardware make supporting new models in on-device inference engines a recurring and costly system challenge. Each model may require coordinated changes across conversion, graph construction, operators, and hardware backends. Existing coding agents often miss these cross-layer dependencies and struggle to diagnose failures across host--device execution paths. We present EngineAgent, an agentic system for automating on-device LLM model enablement. Its key insight is to separate reusable engine and hardware knowledge from model-specific implementation. Before enablement, EngineAgent constructs a hierarchy-aware engine index that captures engine abstraction layers and cross-layer dependencies, and pre-validates backend-specific build, execution, and debugging workflows using already supported models. For each new model, a builder uses this knowledge to coordinate cross-layer changes, while an auditor checks implementation completeness and drives iterative debugging through device-side validation. Across 12 tasks spanning three mobile LLMs and four engine--backend pairs on CPUs, GPUs, and NPUs, EngineAgent achieves 75% end-to-end success with either of two underlying LLMs, versus 33.3% for the strongest coding-agent baseline. Success requires compilation, on-device execution, and output-logit agreement with reference implementations. EngineAgent also reduces average per-task time and model-API cost relative to baselines using the same underlying LLMs.

---

Notes written by AI from the paper, numbers copied from the abstract. Check the paper before you rely on them. From AI Papers Daily.