The paper introduces Inference‑Native Zeroth‑Order (ZO) optimization, which redefines ZO as a query‑based process that can be executed directly by inference runtimes. By exposing ZO’s query semantics and using abstractions such as ProbePlan, factorized side states, and persistent subspace reuse, the method reduces state‑management cost and DRAM traffic dramatically. Experiments on large models (OPT‑13B, Qwen3‑8B) show that inference‑native steps are nearly identical to matched‑query controls while achieving significant memory savings and efficient batching.
arXiv:2609.38090v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models are a compelling architecture for scaling model capacity, making them especially attractive for deployment on resource-...
arXiv:2609.23130v1 Announce Type: new Abstract: Large language model (LLM) inference is evolving from an engine-local optimization problem into a distributed control problem involving reusable state,...
arXiv:2607. 24787v1 Announce Type: new Abstract: Sparse Mixture-of-Experts (MoE) models expand foundation model capacity through conditional expert activation, but their full expert pools remain difficult to deploy under limited accelerator memory.
The paper argues that AI deployment performance depends on interactions among compression, compiler transformations, and serving policies rather than just model architecture. It introduces a three‑layer taxonomy—model‑level techniques, compiler transformations, and system policies—and frames deployment as a constrained multi‑objective optimization problem over accuracy, latency, throughput, memory footprint, and energy. The authors propose an evidence protocol for comparable benchmarking and synthesize data from edge and data‑center platforms to show that cross‑layer interactions drive deployment outcomes, concluding with a constraint‑aware selection procedure and open research problems.
Dynamic Expert Quantization (DynaExq) is a runtime-aware mixed-precision serving system designed for single‑GPU Mixture‑of‑Experts (MoE) inference under a hard high‑bandwidth memory (HBM) envelope. It treats the problem as an online, budget‑constrained precision allocation task, keeping the most frequently used experts at higher precision while relegating the rest to low‑precision fallbacks. By estimating expert hotness from router traces and asynchronously promoting or demoting experts, DynaExq maintains a fully materialized expert set during the forward pass, improving accuracy and throughput compared to static post‑training quantization and offloading/prefetch baselines. whyItMatters":"DynaExq enables efficient deployment of large MoE models on memory‑limited GPUs by dynamically allocating precision based on runtime expert usage, thereby reducing memory footprint and latency while boosting accuracy and throughput."
The paper introduces compositional online learning for semantic data processing systems, addressing the high cost and latency of large language model (LLM) calls. It proposes a framework that combines lightweight online-learning components—such as memoization, per-call filter-ordering, and per-batch cascade-routing—within the LLM call boundary, allowing each component to make real-time decisions and update its models without exceeding the LLM round-trip time. A production case study in Cortex AISQL demonstrates that these components can reduce the per-row LLM cost by up to 8× compared to a baseline workload.
arXiv:2607. 11985v1 Announce Type: cross Abstract: Hybrid quantum-classical machine learning workflows repeatedly evaluate many small parametrized circuits during training and model exploration.
arXiv:2606. 27743v1 Announce Type: cross Abstract: Large Language Models (LLMs) inference is typically deployed under a static resource assumption, where models execute a fixed computational graph regardless of the runtime environment.
arXiv:2608. 08020v1 Announce Type: new Abstract: Test-time compute scaling is a primary driver of performance in large reasoning models (LRMs), but extreme inefficiency bounds current approaches, shifting the critical question from \emph{how much} compute to spend, to \emph{where} to allocate it.
arXiv:2605. 21312v2 Announce Type: replace-cross Abstract: Modern LLM serving is no longer homogeneous or monolithic.
Instella‑MoE is a fully open Mixture‑of‑Experts language model with 16 billion total parameters and 2.8 billion active parameters per token, trained from scratch on AMD Instinct GPUs. It incorporates a sparsely activated MoE design with Gated Multi‑head Latent Attention and FarSkip‑Collective connectivity, and follows a multi‑stage pipeline that includes pre‑training, long‑context extension, supervised fine‑tuning, direct preference optimization, and reinforcement learning with Multi‑Teacher On‑Policy Distillation. The model achieves an average score of 76.7 on pre‑training benchmarks and 73.2 on instruction‑following, reasoning, math, coding, and chat benchmarks, outperforming comparable fully open and open‑weight models, and its full training pipeline, weights, and code are released for reproducibility.
The paper introduces Edge0, a streaming mixture‑of‑experts (MoE) inference engine that enables a 35‑billion‑parameter MoE model to run on consumer hardware by predicting routing decisions one token ahead. Edge0 uses a per‑layer prerouter to prefetch the necessary experts from SSD, and an unmerged recovery LoRA trained on the student path to recover quality lost to 4‑bit quantization and routing replacement. On a single 24‑GB machine, Edge0 serves the 35B MoE at 20 tokens per second while keeping peak active memory below 3 GiB, achieving performance close to its fp16 teacher across five public benchmarks.