Think Before You Grid-Search: Floor-First Triage for LLM Serving

• URL: http://arxiv.org/abs/2607.05876v2
• Date: Wed, 08 Jul 2026 02:47:41 GMT
• Title: Think Before You Grid-Search: Floor-First Triage for LLM Serving
• Abstract summary: LLM serving optimization typically benchmarks many configurations and reaches for heavy profilers when latency targets are missed.<n>We argue for the reverse discipline: estimation is the analytical layer of profiling.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: LLM serving optimization typically benchmarks many configurations and reaches for heavy profilers when latency targets are missed. We argue for the reverse discipline: estimation is the analytical layer of profiling -- without it, optimization degenerates to grid search. Floor First is a residual-driven triage workflow. Each decode step is modeled as a five-dimensional resource vector (HBM bytes, FLOPs, network bytes, network messages, KV capacity); summing within a resource and maximizing across resources gives an optimistic floor, the plain sum a pessimistic one. Where a measurement lands inside this [max, sum] interval reads out overlap quality before any profiler is opened, and profilers escalate only on residuals above a stated threshold. Deployment alternatives are compared by wall ordering -- which resource wall binds first as load grows -- rather than by point benchmarks. The account is compositional: new attention or state-space variants enter by declaring one module, and the workflow ships as a zero-dependency calculator plus an agent skill that enforces the discipline in agentic optimization loops. As a case study we analyze a DeepSeek-V3.2-style 671B MoE/MLA model on 16 NVIDIA H20 GPUs, whose ridge point of ~74 FLOP/byte (vs ~590 for H100) makes it an extreme decode-oriented part. The floors show TP16 decoding is KV-capacity-limited to ~70 concurrent 8K requests; sparse attention removes the KV-bandwidth term but not the capacity wall; an EP16+DP-attention layout accepts slightly worse same-batch weight traffic for an order-of-magnitude higher capacity wall (~644) -- while single-stream latency favors TP by 2.4x. The layout judgment is thus a computable function of the operating point, explaining why production deployments on identical hardware have shipped opposite attention layouts.

Related papers

• From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix [55.08143827481755]
  We consolidate traffic from over 200 internal applications onto a single model.<n>We train a separate GRPO expert per axis and merge them via two-stage SLERP.<n>The model absorbs 50% of platform traffic, 116M requests per month, at a fraction of the serving cost.
  arXiv  Detail & Related papers  (2026-09-01T17:39:26Z)
• Pipeline-Native Transformers: Co-Designing Model Architecture and CPU Inference for Bandwidth-Efficient Autoregressive Decode [0.0]
  Single-token autoregressive decode on CPUs is bound by memory bandwidth, not arithmetic.<n>This report argues that the most effective response is to co-design the model architecture and the inference together.<n>It presents cflow, a CPU-first streaming engine, alongside a family of pipeline-native transformer architectures.
  arXiv  Detail & Related papers  (2026-08-24T21:28:05Z)
• Constrained Decoding for Diffusion Language Models via Efficient Inference over Finite Automata [57.27430779838529]
  We present an exact and tractable algorithm for sampling from the constrained mean-field posterior under any constraint expressible as a finite automaton.<n>The approach guarantees constraint satisfaction by construction, supports both greedy and sampling-based decoding, and is compatible with parallel and block-wise decoding.<n> Empirical evaluations on Dream-7B and LLaDA-8 show substantial accuracy gains across various tasks.
  arXiv  Detail & Related papers  (2026-07-08T05:48:57Z)
• PersistentKV: Page-Aware Decode Scheduling for Long-Context LLM Serving on Commodity GPUs [0.8460698440162889]
  We present PersistentKV, a block-table decode attention engine and page-aware scheduling study.<n>Per PersistentKV maps work by KV-head group, executes directly over native page tables, and adds a compact workqueue schedule.<n>We show that five held-out seeds improve mean wall decode-token throughput by 1.04x to 1.08x on bimodal, uniform, and Zipf-like workloads.
  arXiv  Detail & Related papers  (2026-06-25T06:56:43Z)
• Closed-Form Spectral Regularization for Multi-Task Model Merging [96.82449201305234]
  Model merging combines several independently fine-tuned experts into a single multi-task model without any training data.<n>State-of-the-art merging methods formulate merging as a layer-wise interference problem.<n>We propose SWUDI, a closed-form method that combines a soft exponential filter, which matches the gradient-flow trajectory of iterative descent.
  arXiv  Detail & Related papers  (2026-06-05T14:00:47Z)
• ChunkFT: Byte-Streamed Optimization for Memory-Efficient Full Fine-Tuning [58.54940026861599]
  textscChunkFT is a memory-efficient fine-tuning framework.<n>textscChunkFT enables gradient computation for arbitrary sub-tensors without modifying the network architecture.<n>textscChunkFT consistently outperforms existing memory-efficient baselines.
  arXiv  Detail & Related papers  (2026-05-20T13:44:44Z)
• XFP: Quality-Targeted Adaptive Codebook Quantization with Sparse Outlier Separation for LLM Inference [0.0]
  XFP determines codebook size, outlier budget, and packing per layer automatically.<n>XFP reaches 138 tok/s single-stream decode on workstation hardware.<n>For models that do not fit in the target memory envelope, we present the H-Process.
  arXiv  Detail & Related papers  (2026-05-14T13:52:31Z)
• ZeRO-Prefill: Zero Redundancy Overheads in MoE Prefill Serving [18.574823955774207]
  We propose ZeRO-Prefill, a prefill-only serving system whose backend gathers experts by weight rather than routing them by activation.<n>On Qwen3-235B-A22B across four hardware/precision configurations, ZeRO-Prefill delivers 1.35-1.37x throughput over the strongest distributed baseline on real-world workloads.
  arXiv  Detail & Related papers  (2026-05-03T03:10:24Z)
• GLM-5 Serving Parameter Tuning for OpenClaw: Single-Deployment MaaS Inference Optimization for Long-Context Agent Workloads [6.400220635933459]
  OpenClaw requests are dominated by long, tool-augmented prefixes, including system prompts, conversation history, and tool outputs fed back into the context window.<n>For this workload, with about 28k-30k input tokens and 500 output tokens per request, serving quality is governed by throughput, TTFT, and tail latency rather than short-prompt throughput alone.
  arXiv  Detail & Related papers  (2026-04-27T11:12:33Z)
• Latent-Condensed Transformer for Efficient Long Context Modeling [60.72493959155964]
  We propose Latent-Condensed Attention (LCA) for large language models.<n>LCA condenses context within MLA's latent space, where representation is disentangled into semantic latent vectors and positional keys.<n>LCA achieves up to 2.5$times$ prefilling speedup and 90% KV cache reduction at 128K context.
  arXiv  Detail & Related papers  (2026-04-14T08:40:31Z)
• Resource-Efficient Iterative LLM-Based NAS with Feedback Memory [49.44875022114861]
  Neural Architecture Search (NAS) automates network design, but conventional methods demand substantial computational resources.<n>We propose a closed-loop pipeline leveraging large language models (LLMs) to iteratively generate, evaluate, and refine convolutional neural network architectures.
  arXiv  Detail & Related papers  (2026-03-12T16:00:22Z)
• Nexus:Proactive Intra-GPU Disaggregation of Prefill and Decode in LLM Serving [4.309392302169281]
  Engine-level prefill-decode (PD) disaggregation avoids interference but incurs higher hardware and coordination overhead.<n>PD achieves up to 2.2x higher throughput, 20x lower TTFT, and 2.5x lower TBT than vLLM; outperforms SG by up to 2x; and matches or exceeds disaggregated vLLM.
  arXiv  Detail & Related papers  (2025-07-09T07:27:18Z)
• HiRE: High Recall Approximate Top-$k$ Estimation for Efficient LLM Inference [68.59839755875252]
  HiRE comprises of two novel components: (i) a compression scheme to cheaply predict top-$k$ rows/columns with high recall, followed by full computation restricted to the predicted subset, and (ii) DA-TOP-$k$: an efficient multi-device approximate top-$k$ operator. We demonstrate that on a one billion parameter model, HiRE applied to both the softmax as well as feedforward layers, achieves almost matching pretraining and downstream accuracy, and speeds up inference latency by $1.47times$ on a single TPUv5e device.
  arXiv  Detail & Related papers  (2024-02-14T18:04:36Z)
• KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization [67.74400574357472]
  LLMs are seeing growing use for applications which require large context windows, and with these large context windows KV cache activations surface as the dominant contributor to memory consumption during inference.<n> Quantization is a promising approach for compressing KV cache activations; however, existing solutions fail to represent activations accurately in sub-4-bit precision.<n>Our work, KVQuant, facilitates low precision KV cache quantization by incorporating several novel methods.
  arXiv  Detail & Related papers  (2024-01-31T18:58:14Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.