mlsanityData to 7 Oct 2026

Co-occurrence Patterns of LoRA Adapters in Production Diffusion Model Inference Services

arXiv 2609.23321 ↗First posted 20 Sep 2026cs.DC cs.AI cs.LGcited by 0

New in Diffusion models; no direction yet. Diffusion models

Abstract (arXiv)

Low-rank adaptation (LoRA) has become a key technology for serving large-scale personalized large language models and diffusion models in the cloud. However, the co-occurrence patterns, resource contention relationships, and evolutionary regularities of adapters under production inference workloads have not been systematically or quantitatively studied. Based on GenTD26, Alibaba's production diffusion model inference dataset, this paper adopts a graph-theoretic framework to construct an adapter co-occurrence network and conducts a characterization from both static structure and dynamic evolution. Our main findings are as follows. (1) The co-occurrence network is extremely sparse, and adapter usage frequency follows a significant heavy-tailed distribution. (2) Introducing the first adapter incurs a 66.1% execution-latency overhead, with diminishing marginal costs afterwards. (3) Co-occurrence relationships are driven by base models: in 90.6% of multi-adapter requests, all adapters share the same dominant base model; 66.2% of significant co-occurrence edges connect same-model adapter pairs; and in 85.8% of multi-adapter requests, all adapter pairs form significant co-occurrence edges. (4) The adapter ecosystem exhibits a core-periphery bipolar structure, with a weekly Jaccard similarity of 0.696 at the model level and a churn rate of 54.5% for the top-10 hottest models within a 12-hour window. Based on these findings, we propose a preloading strategy built on top-k co-occurrence statistics; offline experiments show that it covers 81.0% of test-set co-occurrence pairs at k=3, and sensitivity analyses across frequency thresholds and time windows verify the robustness of the conclusions. These results provide a data-driven basis for cache preloading, adaptive scheduling, and GPU memory management in LoRA inference services.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 13 of its 36 references that are in the map. First posted on arXiv 20 Sep 2026.

Builds on

13 of 36 references.

References · 13

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  2. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  3. 2022Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learningcited by 886
  4. 2023Splitwise: Efficient generative LLM inference using phase splittingcited by 536
  5. 2023S-LoRA: Serving Thousands of Concurrent LoRA Adapterscited by 146
  6. 2023Punica: Multi-Tenant LoRA Servingcited by 80
  7. 2019Analysis of Large-Scale Multi-Tenant GPU Clusters for DNN Training Workloadscited by 79
  8. 2023Distributed Inference and Fine-tuning of Large Language Models Over The Internetcited by 42
  9. 2021Characterization and Prediction of Deep Learning Workloads in Large-Scale GPU Datacenterscited by 31
  10. 2024CaraServe: CPU-Assisted and Rank-Aware LoRA Serving for Generative LLM Inferencecited by 21
  11. 2024Chameleon: Adaptive Caching and Scheduling for Many-Adapter LLM Inference Environmentscited by 9
  12. 2025FlexPipe: Adapting Dynamic LLM Serving Through Inflight Pipeline Refactoring in Fragmented Serverless Clusterscited by 9
1 more
  1. 2024SwiftDiffusion: Efficient Diffusion Model Serving with Add-on Modulescited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Serving Heterogeneous LoRA Adapters in Distributed LLM Inference Systems
  2. 2024LoRA-Switch: Boosting the Efficiency of Dynamic LLM Adapters via System-Algorithm Co-design
  3. 2025LoRALib: A Standardized Benchmark for Evaluating LoRA-MoE Methods
  4. 2025LoRA-X: Bridging Foundation Models with Training-Free Cross-Model Adaptation
  5. 2026MinT: Managed Infrastructure for Training and Serving Millions of LLMs
  6. 2025Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA
  7. 2026A Unified Study of LoRA Variants: Taxonomy, Review, Codebase, and Empirical Evaluation
  8. 2024Retrieval-Augmented Mixture of LoRA Experts for Uploadable Machine Learning