AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

CONTINUUM: Restoring the Contiguous Tensor Abstraction Efficiently for Dynamic AI Workloads via Hardware Virtualization

Yangyu Zhang, Shuoming Zhang, Chunwei Xia, Shuaijiang Li, Zhicheng Li, Ruiyuan Xu, Zheming Yang, Lei Chen, YUAN WEN, Guangli Li, Xiaobing Feng, Huimin Cui and 1 more

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace standard CUDA VMM control plane with batched transactional PTE updates in driverGPU VMM mapping · mapping 2 MB pages up to 2GBmapping latency: up to 671.3× speedup, ms to usImproved
use elastic on-demand adapter mapping instead of fixed adapter slotsmulti-LoRA serving vs vLLM · Llama-2-7B, 1 to 16 adaptersthroughput: up to 3.29×Improved
dispatch to cuBLAS on virtually contiguous adapters instead of S-LoRA Triton kernelsmulti-LoRA serving · Llama-2-7Bthroughput speedup: 1.16–1.32×Improved
zero-copy aliasing for tree branching instead of vLLM block-table reconstructiontree decoding · batch size 1, K=8, Lstep=4latency speedup: 7.65×Improved
zero-copy aliasing with token-by-token branchingtree decoding · batch size 1, Lstep=1latency speedup over vLLM: 0.97× to 1.03×No clear effect
increase Lstep (sparser branching) from 4 to 256 (ablation)CONTINUUM tree decoding · batch size 1, K=8latency speedup over vLLM: 7.65× -> 3.64×Hurt
use driver-level VMM backend instead of standard cuMemMap (Ours w/o VMM) (ablation)CONTINUUM · context caching TTFT, 4K-64KTTFTImproved
virtual contiguity with on-demand remapping instead of PyTorch contiguous allocationEngram on DeepSeekMoE-16B config · request rate sweepthroughputImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy