Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912778945560576 |
|---|---|
| author | Li, Allison Greenewald, Kristjan Parnell, Thomas Azizan, Navid |
| author_facet | Li, Allison Greenewald, Kristjan Parnell, Thomas Azizan, Navid |
| contents | Modern large language model (LLM) systems increasingly rely on multi-turn pipelines that are composed of multiple task-specific adapters, yet existing serving frameworks remain inefficient, incurring substantial recomputation overhead when switching between adapters. We present the first LLM serving engine that supports cross-model prefix cache reuse between base and adapted models via Activated LoRA (aLoRA), enabling efficient and fine-grained adapter switching during inference. Our design extends the vLLM framework by introducing base-aligned block hashing and activation-aware masking within the model execution path, permitting cache reuse across models while preserving compatibility with existing serving engine optimizations. Integrated into a production-grade inference stack, this approach supports dynamic adapter activation without excessive key-value tensor recomputation. Evaluation across representative multi-turn, multi-adapter pipelines demonstrates up to 58x end-to-end latency reduction and over 100x time-to-first-token improvement relative to standard LoRA baselines, with benefits that scale with model size and sequence length and manifest across all stages of the request lifecycle. This work bridges parameter-efficient model adaptation with high-performance serving, providing the first complete realization of cross-model KV-cache reuse in modern LLM inference engines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_17910 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA Li, Allison Greenewald, Kristjan Parnell, Thomas Azizan, Navid Distributed, Parallel, and Cluster Computing Artificial Intelligence Machine Learning Modern large language model (LLM) systems increasingly rely on multi-turn pipelines that are composed of multiple task-specific adapters, yet existing serving frameworks remain inefficient, incurring substantial recomputation overhead when switching between adapters. We present the first LLM serving engine that supports cross-model prefix cache reuse between base and adapted models via Activated LoRA (aLoRA), enabling efficient and fine-grained adapter switching during inference. Our design extends the vLLM framework by introducing base-aligned block hashing and activation-aware masking within the model execution path, permitting cache reuse across models while preserving compatibility with existing serving engine optimizations. Integrated into a production-grade inference stack, this approach supports dynamic adapter activation without excessive key-value tensor recomputation. Evaluation across representative multi-turn, multi-adapter pipelines demonstrates up to 58x end-to-end latency reduction and over 100x time-to-first-token improvement relative to standard LoRA baselines, with benefits that scale with model size and sequence length and manifest across all stages of the request lifecycle. This work bridges parameter-efficient model adaptation with high-performance serving, providing the first complete realization of cross-model KV-cache reuse in modern LLM inference engines. |
| title | Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA |
| topic | Distributed, Parallel, and Cluster Computing Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2512.17910 |