Back to AI News
arxiv

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

EngineAgent is an agentic system that automates enabling new LLM models in on-device inference engines by separating reusable engine and hardware knowledge from model-specific implementation. It builds a hierarchy-aware engine index, pre-validates backend workflows, and uses a builder-auditor loop to coordinate cross-layer changes and device-side debugging. Across 12 tasks covering three mobile LLMs and four engine-backend pairs on CPUs, GPUs, and NPUs, it achieved 75% end-to-end success versus 33.3% for the strongest coding-agent baseline, while reducing average per-task time and model-API cost.

AI Brief

It presents a concrete agentic approach with measured gains for the costly engineering problem of bringing new LLMs onto heterogeneous mobile inference stacks.

arXiv:2611.14356v1 Announce Type: cross 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.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!