Publication Radar · Source metadata

LoRazor: Trimming Adapter Execution for Efficient Multi-LoRA Serving

Xiang Li, Pengcheng Wang, Azam Ikram, Saurabh Bagchi

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First announced by arXiv: 2026-11-10 · Indexed by Pub-AI: 2026-11-10 UTC

preprint

LoRazor: Trimming Adapter Execution for Efficient Multi-LoRA Serving

Xiang Li, Pengcheng Wang, Azam Ikram, Saurabh Bagchi

First announced by arXiv: 2026-11-10 · Original record ↗

arxiv
2611.10049
Explicit source versions (1)
  • v1 · Date not supplied
  1. Preprint available

    Source date: 2026-11-10 · First detected: 2026-11-10 UTC · Historical baseline

    Source evidence (1)
    • Official evidence ↗ · RSS guid=oai:arXiv.org:2611.10049v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv

LoRA offers a parameter-efficient adaptation strategy, supporting multi-user workloads via customized adapters on a shared backbone. However, serving concurrent requests with different adapters introduces substantial computation and memory-management overhead. We ask the research question: Do all requests need their adapters throughout the whole generation process? Our motivational study reveals that employing LoRA adapters at every decoding step is unnecessary, releasing them early has limited impact on accuracy while substantially accelerating generation. Although skipping adapters early offers theoretical workload reduction, translating these computational savings into concrete throughput gains requires tailored GPU kernels. We present LoRazor, a multi-LoRA serving framework that combines early adapter release with active-row aware GPU kernels. LoRazor determines when to permanently release a request's adapter and continue decoding with the backbone alone. By dynamically resizing LoRA kernel launches as adapters are released, LoRazor trims idle GPU blocks. Across two backbones and four tasks, LoRazor achieves up to 3.02x request throughput and 2.53x output-token throughput over vLLM multi-LoRA, with a mean relative task-accuracy decrease of 2.89%.