LLM Inference

LLM: Large Language Model

Follow this topic Ask about this topic

Large language model inference is the process of turning prompts into outputs, and its speed, cost, reliability, and privacy shape what applications can use models in practice. Systems work optimizes prefill and token-by-token decoding through KV-cache management, batching and scheduling, parallelism, quantization, and specialized GPU, mobile, or edge kernels. Another strand changes generation at test time using speculative decoding, early exit, retrieval, or repeated reasoning paths to trade extra computation for better accuracy or confidence. The central challenge is that these trade-offs depend heavily on workload, hardware, and inference backend, making reproducible benchmarks and methods that preserve quality while reducing latency, memory, and energy ongoing priorities.

Latest papers

285 papers44 papers in the last four weeks, up 63% on the four weeks before. 0.3% of all new papers.

Mar 27, 2026

Accelerating PayPal's Commerce Agent with Speculative Decoding: An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models

We evaluate speculative decoding with EAGLE3 as an inference-time optimization for PayPal's Commerce Agent, powered by a fine-tuned llama3.1-nemotron-nano-8B-v1 model. Building on prior work (NEMO-4-PAYPAL) that reduced latency and cost through domain-specific fine-tuning, we benchmark EAGLE3 via vLLM against NVIDIA NIM on identical 2xH100 hardware across 40 configurations spanning speculative token counts (gamma=3, gamma=5), concurrency levels (1-32), and sampling temperatures (0, 0.5). Key findings: (1) gamma=3 achieves 22-49% throughput improvement and 18-33% latency reduction at zero additional hardware cost; (2) acceptance rates remain stable at approximately 35.5% for gamma=3 across all conditions; (3) gamma=5 yields diminishing returns (approximately 25% acceptance rate); (4) LLM-as-Judge evaluation confirms fully preserved output quality; and (5) speculative decoding on a single H100 matches or exceeds NIM on two H100s, enabling 50% GPU cost reduction.

Ally Qin, Jian Wan, Sarat Mudunuri +1

LLM Inference LLM Inference Acceleration Speculative Decoding

Mar 26, 2026

GhostServe: A Lightweight Checkpointing System in the Shadow for Fault-Tolerant LLM Serving

The rise of million-token, agent-based applications has placed unprecedented demands on large language model (LLM) inference services. The long-running nature of these tasks increases their susceptibility to hardware and software faults, leading to costly job failures, wasted resources, and degraded user experience. The stateful key-value (KV) cache, which grows with the sequence length, presents a central challenge as it is a critical and vulnerable component in distributed serving systems. In this work, we propose GhostServe, a novel checkpointing solution to facilitate fault-tolerant LLM serving. Specifically, GhostServe protects the streaming KV cache in the shadow by applying erasure coding to generate and store the parity shards in host memory. In the event of device failures, GhostServe enables fast reconstruction of the lost KV cache, allowing the inference process to resume seamlessly without costly full recomputation or state replication. Evaluations demonstrate that GhostServe reduces checkpointing latency by up to 2.7x and recovery latency by 2.1x for a single batch, and 1.2x median response latency compared to existing methods, in the presence of system failures, paving the way for high-availability and cost-effective LLM serving at scale.

Shakya Jayakody, Youpeng Zhao, Chinmay Dhanraj Nehate +1

University of Central Florida, Orlando, FL, USA.

LLM Serving LLM Inference KV-Cache Management

Mar 25, 2026

Secure On-Premise Deployment of Open-Weights Large Language Models in Radiology: An Isolation-First Architecture with Prospective Pilot Evaluation

Purpose: To design, implement, evaluate, and report on the regulatory requirements of a self-hosted LLM infrastructure for radiology adhering to the principle of least privilege, emphasizing technical feasibility, network isolation, and clinical utility. Materials and Methods: The isolation-first, containerized LLM inference stack relies on strict network segmentation, host-enforced egress filtering, and active isolation monitoring preventing unauthorized external connectivity. An accompanying deployment package provides automated isolation and hardening tests. The system served the open-weights DeepSeek-R1 model via vLLM. In a one-week pilot phase, 22 residents and radiologists were free to use 10 predefined prompt-templates whenever they considered them useful in daily work. Afterward, they rated clinical utility and system stability on an 0-10 Likert scale and reported observed critical errors in model output. Results: The applied institutional governance pathway achieved approval from clinic management, compliance, data protection and information security officers for processing unanonymized PHI. The system was rated stable and user friendly during the pilot. Source text-anchored tasks, such as report corrections or simplifications, and radiology guideline recommendations received the highest utility ratings, whereas open-ended conclusion generation based on findings resulted in the highest frequency of critical errors, such as clinically relevant hallucinations or omissions. Conclusion: The proposed isolation-first on-premise architecture enabled overcoming regulatory borders, showed promising clinical utility in text-anchored tasks and is the current base to serve open-weights LLMs as an official service of a German University Hospital with over 10,000 employees. The deployment package were made publicly available (https://github.com/ukbonn/ukb-gpt).

Sebastian Nowak, Jann-Frederick Laß, Narine Mesropyan +8

Department of Diagnostic and Interventional Radiology, University Hospital Bonn · Department of Research-IT, University Hospital Bonn

LLM Serving LLM Security LLM Inference

Mar 24, 2026

Energy Efficient Software Hardware CoDesign for Machine Learning: From TinyML to Large Language Models

The rapid deployment of machine learning across platforms from milliwatt-class TinyML devices to large language models has made energy efficiency a primary constraint for sustainable AI. Across these scales, performance and energy are increasingly limited by data movement and memory-system behavior rather than by arithmetic throughput alone. This work reviews energy efficient software hardware codesign methods spanning edge inference and training to datacenter-scale LLM serving, covering accelerator architectures (e.g., ASIC/FPGA dataflows, processing-/compute-in-memory designs) and system-level techniques (e.g., partitioning, quantization, scheduling, and runtime adaptation). We distill common design levers and trade-offs, and highlight recurring gaps including limited cross-platform generalization, large and costly co-design search spaces, and inconsistent benchmarking across workloads and deployment settings. Finally, we outline a hierarchical decomposition perspective that maps optimization strategies to computational roles and supports incremental adaptation, offering practical guidance for building energy and carbon aware ML systems.

Mohammad Saleh Vahdatpour, Yanqing Zhang

Department of Computer Science Georgia State University Atlanta, GA, USA

Power-Efficient Inference Energy-Efficient ML Hardware-Software Co-Design LLM Inference Tiny ML

Mar 22, 2026

TIDE: Token-Informed Depth Execution for Per-Token Early Exit in LLM Inference

Large language models run every token through every layer, regardless of difficulty. We present TIDE, a post-training system that attaches tiny learned routers at periodic checkpoint layers and, at inference time, selects the earliest layer whose hidden state has converged for each token. TIDE requires no model retraining, works with any HuggingFace causal LM, auto-detects GPU architecture, and supports float32, float16, and bfloat16 through fused CUDA kernels. On an NVIDIA A100 with DeepSeek R1 Distill 8B, TIDE achieves 100% prefill exit rate (5% of tokens exit at layer 11, the remaining at layer 31), reduces prefill latency by 7.2%, and increases single-batch throughput by 6.6%. During autoregressive decoding, 98-99% of tokens exit early while the model correctly solves a multi-step math problem with 95 unique output tokens. On Qwen3 8B (36 layers), throughput improves by 8.1% at batch size 8. Calibration on 2,000 WikiText samples takes under 3 minutes and produces a ~4 MB router checkpoint. The system comprises 1,308 lines of Python and 1,081 lines of CUDA/C++ with 74 passing tests. Code: https://github.com/RightNow-AI/TIDE

Jaber Jaber, Osama Jaber

RightNow AI

LLM Inference LLM Inference Acceleration Layer Skipping in Language Model Decoding

Mar 22, 2026

Enhancing reasoning accuracy in large language models during inference time

Large Language Models (LLMs) often exhibit strong linguistic abilities while remaining unreliable on multi-step reasoning tasks, particularly when deployed without additional training or fine-tuning. In this work, we study inference-time techniques to improve the reasoning accuracy of LLMs. We systematically evaluate three classes of inference-time strategies: (i) self-consistency via stochastic decoding, where the model is sampled multiple times using controlled temperature and nucleus sampling and the most frequent final answer is selected; (ii) dual-model reasoning agreement, where outputs from two independent models are compared and only consistent reasoning traces are trusted; and (iii) self-reflection, where the model critiques and revises its own reasoning. Across all evaluated methods, we employ Chain-of-Thought (CoT) [1] prompting to elicit explicit intermediate reasoning steps before generating final answers. In this work, we provide a controlled comparative evaluation across three inference-time strategies under identical prompting and verification settings. Our experiments on LLM [2] show that self-consistency with nucleus sampling and controlled temperature value yields the substantial gains, achieving a 9% to 15% absolute improvement in accuracy over greedy single-pass decoding, well-suited for low-risk domains, offering meaningful gains with minimal overhead. The dual-model approach provides additional confirmation for model reasoning steps thus more appropriate for moderate-risk domains, where higher reliability justifies additional compute. Self-reflection offers only marginal improvements, suggesting limited effectiveness for smaller non-reasoning models at inference time.

Vinay Sharma, Manish Jain

LLM Inference LLM Reasoning Inference-Time Guidance Self-Consistency Decoding

Mar 17, 2026

NanoZK: Privacy-Preserving Verifiable Inference for Large Language Models via Layerwise Zero-Knowledge Proofs

We present NanoZK, a zero-knowledge proof system for verifiable LLM inference: clients and third-party auditors check that a provider executed the advertised model on a committed input without learning weights or activations. NanoZK introduces a layerwise proof framework that decomposes transformer inference into independently provable layers linked by a SHA-256 commitment chain, yielding constant-size sub-circuit proofs (3.5-3.7 KB; about 83 KB total at L=12), comparable in total size to and substantially more parallelizable than prior ZKML's monolithic 101-126 KB proofs. We prove compositional soundness and zero-knowledge under standard assumptions, design 16-bit lookup-table approximations for softmax, GELU, and normalization with measured perplexity degradation below 1e-4 across six model/dataset combinations, and add a Fisher-information-guided audit-budget triage as an efficiency tool (full soundness still requires verifying every layer). On CPU the MLP sub-circuit proves in about 6.3 s prove-only (about 43 s setup plus prove) with about 22 ms verification at any width; attention prove-only time scales from 0.9 s (d=16) to 184 s (d=256); full-block end-to-end proofs are measured to d=128, with a projected GPU time of about 68 s per block at d=768 from measured O(d^2) MSM scaling and a conservative 15-30x GPU-MSM speedup range based on Icicle's published 30x result for n >= 2^20 and extrapolated to the smaller-n regime. Privacy scope: NanoZK hides weights and activations from verifiers and auditors but does not hide the prompt from the prover; this is complementary to HE/MPC.

Zhaohui Wang

USC Viterbi School of Engineering, Los Angeles, CA, USA

Zero-Knowledge ML Privacy-Preserving ML Privacy-Preserving Language Models LLM Inference

Mar 16, 2026

This Is Taking Too Long -- Investigating Time as a Proxy for Energy Consumption of LLMs

The energy consumption of Large Language Models (LLMs) is raising growing concerns due to their adverse effects on environmental stability and resource use. Yet, these energy costs remain largely opaque to users, especially when models are accessed through an API -- a black box in which all information depends on what providers choose to disclose. In this work, we investigate inference time measurements as a proxy to approximate the associated energy costs of API-based LLMs. We ground our approach by comparing our estimations with actual energy measurements from locally hosted equivalents. Our results show that time measurements allow us to infer GPU models for API-based LLMs, grounding our energy cost estimations. Our work aims to create means for understanding the associated energy costs of API-based LLMs, especially for end users.

Lars Krupp, Daniel Geißler, Francisco M. Calatrava-Nicolas +3

Embedded Intelligence DFKI and RPTU Kaiserslautern, Germany · Centre for Applied Autonomous Sensor Systems (AASS) Örebro University Örebro, Sweden

Energy-Aware LLM Inference LLM Inference Energy-Efficient ML

Mar 14, 2026

Clinician input steers AI toward accurate and harmful recommendations

Large language models (LLMs) are entering clinical workflows, yet evaluations rarely assess how clinician reasoning shapes model behavior during clinical interactions. Using 61 curated NEJM Case Records, we tested how expert or misleading clinician reasoning influenced AI-generated differential diagnoses and next step recommendations across 21 reasoning variants from 8 proprietary and open-source models. After clinician exposure, LLM-clinician concordance increased: simulations with >=3 overlapping differential diagnoses rose from 65.8% to 93.5%, and those with >=3 overlapping next step recommendations from 20.3% to 53.8%. Expert context significantly improved correct final-diagnosis inclusion in all 21 models (mean +20.4 pp), reflecting both improved reasoning and passive content echoing, while adversarial context significantly degraded performance in 14 models (mean -5.4 pp). Expert context also significantly increased leading-diagnosis accuracy in all 21 models, whereas adversarial context significantly reduced it in 13. Multi-turn disagreement challenges revealed distinct model phenotypes, from highly conformist to dogmatic, with adversarial arguments remaining a vulnerability even in otherwise resilient models. Inference-time scaling reduced harmful echoing of clinician-introduced recommendations across WHO harm-severity tiers by 62.7% for mild, 57.9% for moderate, 76.3% for severe, and 83.5% for death-tier recommendations. Inference-time prompting recovered diagnostic accuracy lost to adversarial context while preserving expert-context benefits across GPT-5, Claude Sonnet 4.5, and Gemini 3 Flash, and sharply reduced highly consistent harmful echoing across severity tiers. These findings provide a foundation for evaluating clinician-AI collaboration and introduce interactive metrics and mitigation strategies essential to safety and robustness.

Ivan Lopez, Selin S. Everett, Bryan J. Bunning +10

Department of Biomedical Data Science, Stanford University, Stanford, CA, USA · Stanford University School of Medicine, Stanford, CA, USA · Quantitative Sciences Unit, Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA +14

Medical LLMs Clinical Language Model Evaluation LLM Sycophancy Adversarial Robustness of Language Models LLM Inference Clinical Reasoning

Mar 14, 2026

Domain-Specialized Tree of Thought through Plug-and-Play Predictors

While Large Language Models (LLMs) have advanced complex reasoning, prominent methods like the Tree of Thoughts (ToT) framework face a critical trade-off between exploration depth and computational efficiency. Existing ToT implementations often rely on heavyweight LLM-based self-evaluation or rigid heuristics for branch pruning, making them prohibitively expensive and inflexible for broad application. To address this, we introduce DST, an adaptable, plug-and-play predictor that serves as a lightweight, supervised heuristic to guide the ToT search process. Our predictor enables dynamic, context-aware pruning, allowing the search to proceed with near-greedy efficiency on simpler reasoning steps while adaptively expanding the search beam only when encountering uncertainty or task complexity. We evaluate our approach on a diverse suite of benchmarks spanning mathematical reasoning, general reasoning, and complex logical reasoning. Experimental results demonstrate that our method achieves accuracy competitive with or superior to strong baselines, including standard ToT, while reducing computational overhead by 26-75%. Our work effectively resolves the accuracy-efficiency trade-off in tree-based reasoning, transforming ToT from a resource-intensive technique into a scalable and practical paradigm for complex problem-solving in LLMs.

Xuanqi Gao, Haoyu Wang, Jun Sun +2

Xi’an Jiaotong University, China · Singapore Management University, Singapore · University of Massachusetts at Amherst, United States

LLM Reasoning Tree of Thoughts Efficient Language Model Reasoning LLM Inference

Mar 14, 2026

Intelligent Materials Modelling: Large Language Models Versus Partial Least Squares Regression for Predicting Polysulfone Membrane Mechanical Performance

Predicting the mechanical properties of polysulfone (PSF) membranes from structural descriptors remains challenging due to extreme data scarcity typical of experimental studies. To investigate this issue, this study benchmarked knowledge-driven inference using four large language models (LLMs) (DeepSeek-V3, DeepSeek-R1, ChatGPT-4o, and GPT-5) against partial least squares (PLS) regression for predicting Young's modulus (E), tensile strength (TS), and elongation at break (EL) based on pore diameter (PD), contact angle (CA), thickness (T), and porosity (P) measurements. These knowledge-driven approaches demonstrated property-specific advantages over the chemometric baseline. For EL, LLMs achieved statistically significant improvements, with DeepSeek-R1 and GPT-5 delivering 40.5% and 40.3% of Root Mean Square Error reductions, respectively, reducing mean absolute errors from

11.63\pm5.34

% to

5.18\pm0.17

%. Run-to-run variability was markedly compressed for LLMs (

\leq

3%) compared to PLS (up to 47%). E and TS predictions showed statistical parity between approaches (

q\geq0.05

), indicating sufficient performance of linear methods for properties with strong structure-property correlations. Error topology analysis revealed systematic regression-to-the-mean behavior dominated by data-regime effects rather than model-family limitations. These findings establish that LLMs excel for non-linear, constraint-sensitive properties under bootstrap instability, while PLS remains competitive for linear relationships requiring interpretable latent-variable decompositions. The demonstrated complementarity suggests hybrid architectures leveraging LLM-encoded knowledge within interpretable frameworks may optimise small-data materials discovery.

Dingding Cao, Mieow Kee Chan, Wan Sieng Yeo +2

School of Big Data, Baoshan University, Baoshan, 678000, Yunnan Province, China. · Centre for Water Research, Faculty of Engineering, Built Environment and Information Technology, SEGi University. Jalan Teknologi, Kota Damansara, 47810 Petaling Jaya, Selangor Darul Ehsan, Malaysia. · School of Energy and Chemical Engineering, Xiamen University Malaysia, Jalan Sunsuria, Bandar Sunsuria, 43900 Sepang, Selangor, Malaysia. +3

Materials Property Prediction Polymer Property Prediction Materials Science LLM Inference Linear Regression

Mar 12, 2026

TaxBreak: Unmasking the Hidden Costs of LLM Inference Through Overhead Decomposition

Large Language Model (LLM) inference is widely used in interactive assistants and agentic systems. In latency-sensitive deployments, inference time can become dominated by host-side overheads. Existing approaches typically expose this cost only as an aggregate residual or a launch/queue metric, which is often insufficient to identify which execution layer should be optimized. This work presents TaxBreak, a trace-driven methodology for decomposing host-visible orchestration overhead into three components: framework translation time, CUDA library translation time, and kernel launch-path time. We validate TaxBreak on NVIDIA H100 and H200 systems and use it to derive our proposed Host-Device Balance Index (HDBI), a boundedness summary index that relates device-active execution to host-visible orchestration. Across representative dense and mixture-of-experts workloads in both prefill and decode, we show that aggregate latency, GPU inactivity, or boundedness ratios alone can obscure the dominant optimization target. TaxBreak instead distinguishes cases where optimization should reduce software-stack overhead from cases where the primary win comes from reducing device-side work. We further show that MoE models dispatch 8-11x more kernels per output token than dense models, and that for such host-bound workloads, CPU single-thread performance is a first-order parameter: a faster host CPU reduces orchestration overhead by 10-29% and improves end-to-end latency by up to 14%, even when paired with a slower-clocked GPU. These results position TaxBreak as a diagnostic tool for assessing whether optimization effort should target the software stack or the device-side workload execution.

Prabhu Vellaisamy, Shreesh Tripathi, Vignesh Natarajan +3

NeuroAI Computer Architecture Lab (NCAL), Electrical and Computer Engineering Department, Carnegie Mellon University

LLM Inference Efficiency LLM Inference

Mar 12, 2026

Characterizing Performance-Energy Trade-offs of Large Language Models in Multi-Request Workflows

Large language models (LLMs) are increasingly used in applications forming multi-request workflows like document summarization, search-based copilots, and multi-agent programming. While these workflows unlock richer functionality, they also amplify latency and energy demand during inference. Existing measurement and benchmarking efforts either focus on assessing LLM inference systems or consider single-request evaluations, overlooking workflow dependencies and cross-request interactions unique to multi-request workflows. Moreover, the energy usage of such interdependent LLM calls remains underexplored. To address these gaps, this paper presents the first systematic characterization of performance-energy trade-offs in multi-request LLM inference. We develop four representative workloads capturing sequential, interactive, agentic, and composite patterns common in modern deployments. Using an NVIDIA A100 testbed with state-of-the-art serving systems (vLLM and Parrot), we analyze how key energy knobs affect latency, throughput, and component-level energy use. Our findings reveal batch size as the most impactful lever, though benefits are workload dependent. While optimal batching benefits workloads with large shared prompts, it is ineffective for sequential summarization and only partially effective for multi-agent coding. GPU power capping provides modest but predictable savings, while output length induces linear energy scaling with limited efficiency gains. We further show that engine-level optimizations in vLLM maintain higher GPU utilization and efficiency, especially for decode-heavy workloads, while Parrot's workflow-aware scheduling achieves lower energy consumption under strict power constraints. These findings offer actionable guidelines for developers and system operators designing performance- and energy-aware LLM serving systems in emerging multi-request workflows.

Md. Monzurul Amin Ifath, Israat Haque

Dalhousie University, Halifax, NS, Canada

Parallel LLM Inference LLM Inference LLM Serving Energy-Efficient ML

Mar 12, 2026

BLooP: Zero-Shot Abstractive Summarization using Large Language Models with Bigram Lookahead Promotion

Abstractive summarization requires models to generate summaries that convey information in the source document. While large language models can generate summaries without fine-tuning, they often miss key details and include extraneous information. We propose BLooP (Bigram Lookahead Promotion), a simple training-free decoding intervention that encourages large language models (LLMs) to generate tokens that form bigrams from the source document. BLooP operates through a hash table lookup at each decoding step, requiring no training, fine-tuning, or model modification. We demonstrate improvements in ROUGE and BARTScore for Llama-3.1-8B-Instruct, Mistral-Nemo-Instruct-2407, and Gemma-2-9b-it on CNN/DM, CCSum, Multi-News, and SciTLDR. Human evaluation shows that BLooP significantly improves faithfulness without reducing readability. We make the code available at https://github.com/varuniyer/BLooP

Varun Iyer, Cornelia Caragea

University of Illinois Chicago

Text Summarization Language Model Decoding LLM Inference

Mar 11, 2026

Markovian Generation Chains in Large Language Models

The widespread use of large language models (LLMs) raises an important question: how do texts evolve when they are repeatedly processed by LLMs? In this paper, we define this iterative inference process as Markovian generation chains, where each step takes a specific prompt template and the previous output as input, without including any prior data. In iterative rephrasing and round-trip translation experiments, the output either converges to a small recurrent set or continues to produce novel sentences within a finite horizon. Through sentence-level Markov chain modeling and analysis of simulated data, we show that iterative process can either increase or reduce the diversity depending on factors such as the temperature parameter and the initial input sentence. These results offer valuable insights into the dynamics of iterative LLM inference and their implications for multi-agent LLM systems.

Mingmeng Geng, Amr Mohamed, Guokan Shang +2

ENS-PSL & CNRS-Lattice · MBZUAI & Ecole Polytechnique · MBZUAI

LLM Inference Diverse Text Generation Markov Chains Markov Models

Mar 10, 2026

Summarize Before You Speak with ARACH: A Training-Free Inference-Time Plug-In for Enhancing LLMs via Global Attention Reallocation

Large language models (LLMs) achieve remarkable performance, yet further gains often require costly training. This has motivated growing interest in post-training techniques-especially training-free approaches that improve models at inference time without updating weights. Most training-free methods treat the model as a black box and improve outputs via input/output-level interventions, such as prompt design and test-time scaling through repeated sampling, reranking/verification, or search. In contrast, they rarely offer a plug-and-play mechanism to intervene in a model's internal computation. We propose ARACH(Attention Reallocation via an Adaptive Context Hub), a training-free inference-time plug-in that augments LLMs with an adaptive context hub to aggregate context and reallocate attention. Extensive experiments across multiple language modeling tasks show consistent improvements with modest inference overhead and no parameter updates. Attention analyses further suggest that ARACH mitigates the attention sink phenomenon. These results indicate that engineering a model's internal computation offers a distinct inference-time strategy, fundamentally different from both prompt-based test-time methods and training-based post-training approaches.

Jingtao Wang, Yucong Wang, Jun Ding +2

Meakins-Christe Laboratories, Research Institute of McGill University Health Centre, 1001 Decarie Blvd, Montreal, H4A 3J1, Quebec, Canada · Department of Medicine, Division of Experimental Medicine, McGill University, 1001 Decarie Blvd, Montreal, H4A 3J1, Quebec, Canada · The College of Computer Science and Technology, Zhejiang Gongshang University, Hangzhou, China +4

LLM Inference Inference-Time Adaptation of Language Models Inference-Time Optimization

Mar 9, 2026

Can LLMs Perceive Time? An Empirical Investigation

Large language models cannot estimate how long their own tasks take. We investigate this limitation through four experiments across 68 tasks and four model families. Pre-task estimates overshoot actual duration by 4--7

\times

(

p < 0.001

), with models predicting human-scale minutes for tasks completing in seconds. Relative ordering fares no better: on task pairs designed to expose heuristic reliance, models score at or below chance (GPT-5: 18% on counter-intuitive pairs,

p = 0.033

), systematically failing when complexity labels mislead. Post-hoc recall is disconnected from reality -- estimates diverge from actuals by an order of magnitude in either direction. These failures persist in multi-step agentic settings, with errors of 5--10

\times

. The models possess propositional knowledge about duration from training but lack experiential grounding in their own inference time, with practical implications for agent scheduling, planning and time-critical scenarios.

Aniketh Garikaparthi

TCS Research

Temporal Reasoning in Language Models LLM Inference LLM Reliability

Mar 9, 2026

Reject, Resample, Repeat: Understanding Parallel Reasoning in Language Model Inference

Inference-time methods that aggregate and prune multiple samples have emerged as a powerful paradigm for steering large language models, yet we lack any principled understanding of their accuracy-cost tradeoffs. In this paper, we introduce a route to rigorously study such approaches using the lens of particle filtering algorithms such as Sequential Monte Carlo (SMC). Given a base language model and a process reward model estimating expected terminal rewards, we ask: how accurately can we sample from a target distribution given some number of process reward evaluations? Theoretically, we identify (1) simple criteria enabling non-asymptotic guarantees for SMC; (2) algorithmic improvements to SMC; and (3) a fundamental limit faced by all particle filtering methods. Empirically, we demonstrate that our theoretical criteria effectively govern the sampling error of SMC, though not necessarily its final accuracy, suggesting that theoretical perspectives beyond sampling may be necessary.

Noah Golowich, Fan Chen, Dhruv Rohatgi +4

Microsoft Research · MIT · NYU

Parallel Reasoning in Language Models LLM Inference Parallel LLM Inference Inference-Time Guidance Efficient Language Model Reasoning

Mar 8, 2026

Trusting What You Cannot See: Auditable Fine-Tuning and Inference for Proprietary AI

Cloud-based infrastructure has become the dominant platform for deploying large models, particularly large language models (LLMs). Fine-tuning and inference are increasingly delegated to cloud providers for simplified deployment and access to proprietary models, yet this creates a fundamental trust gap. Although cryptographic and TEE-based verification approaches exist, prohibitive proving costs and limited TEE memory prevent them from scaling to modern LLMs, leaving clients unable to practically audit these processes. This lack of transparency creates concrete security risks that can silently compromise service integrity. We present AFTUNE, an auditable and verifiable framework that ensures the computational integrity of cloud-based fine-tuning and inference. AFTUNE incorporates a lightweight recording and spot-check mechanism that produces verifiable traces of execution. These traces enable clients to later audit whether the fine-tuning and inference processes followed the agreed configurations, by verifying sampled execution blocks inside a TEE, each covering only a small portion of the model and the execution trace. Our evaluation shows that AFTUNE adds modest overhead and makes auditing practical for clients.

Heng Jin, Chaoyu Zhang, Hexuan Yu +4

Virginia Tech · Washington University in St. Louis

Verifiable Outsourced Computation LLM Auditing LLM Inference LLM Fine-Tuning

Mar 6, 2026

M2K: Making the Model-Kernel Interface Explicit for Reliable CUDA Kernel Verification

Large language model (LLM) inference systems rely on CUDA kernels for core GPU computations, yet the interface between models and kernels is implicit and poorly specified. Models and kernels evolve independently and often make incompatible assumptions about tensor shapes and input sizes, leading to subtle memory bugs in CUDA kernels. These bugs can crash inference services, corrupt model weights, or be exploited by remote adversaries. Existing techniques either incur prohibitive runtime overhead, require specialized hardware, or fail to handle dynamic tensor shapes and variable kernel launch configurations, leaving the CUDA memory bugs largely unaddressed. This paper presents M2K, a fully automated framework that makes the model-kernel interface explicit and leverages it to detect memory bugs in CUDA kernels used in LLM inference systems. M2K consists of two components. HFProbe traces model execution without GPU hardware, classifies kernel arguments into model-fixed and user-variable, and emits symbolic constraints that capture the interface. cuKLEE then performs symbolic execution on CUDA kernels to pinpoint memory bugs under the interface constraints, modeling tensors as disjoint memory regions and treating thread identifiers symbolically to scale to thousands of threads. In the evaluation, M2K discovers 181 previously unknown bugs in real LLM inference systems, while producing only nine false positives, demonstrating its effectiveness.

Mengting He, Shihao Xia, Haomin Jia +2

Penn State, USA · SKLP, ICT, CAS, China · Peking University, China

GPU Kernel Correctness Verification LLM Inference

Mar 5, 2026

Parallelization Strategies for Dense LLM Deployment: Navigating Through Application-Specific Tradeoffs and Bottlenecks

Breakthroughs in the generative AI domain have fueled an explosion of large language model (LLM)-powered applications, whose workloads fundamentally consist of sequences of inferences through transformer architectures. Within this rapidly expanding ecosystem, dense LLMs--those that activate all model parameters for each token generation--form the foundation for advanced expert-based variants. Dense models continue to dominate because of their strong generalization ability, scalability, ease of fine-tuning, and versatility across diverse tasks. In LLM inference systems, performance is mainly characterized by latency, response time, and throughput (i.e., tokens generated per unit of time). Latency and throughput are inherently coupled: optimizing for one often comes at the expense of the other. Moreover, batching strategies and parallelism configurations, which are essential when dense model parameters exceed device memory capacity, can significantly affect both latency and overall system throughput. This paper (i) investigates the workloads of two representative dense LLMs--Llama-3.1-70B and Llama-3.1-405B, focusing in particular on intra-node parallelization schemes, (ii) analyzes how input characteristics, batching, and parallelism strategies influence latency flexibility and the latency-throughput tradeoff, and (iii) identifies key performance bottlenecks that inform design choices for meeting service-level agreements (SLAs) and sustaining inference quality. Our empirical evaluations reveal that Tensor Parallelism (TP) improves the latency objectives while Pipeline Parallelism (PP) is better-suited for throughput-oriented applications. We highlight that their hybrid usage by controlling the TP and PP degrees provides control over the latency-throughput interplay.

Burak Topcu, Musa Oguzhan Cim, Poovaiah Palangappa +2

The Pennsylvania State University, State College, Pennsylvania, USA · Advanced Micro Devices, San Jose, California, USA · Advanced Micro Devices, Portland, Oregon, USA

Parallel LLM Inference LLM Inference Tensor Parallelism Pipeline Parallelism

Mar 4, 2026

From Static Inference to Dynamic Interaction: A Survey of Streaming Large Language Models

Standard Large Language Models (LLMs) are predominantly designed for static inference with pre-defined inputs, which limits their applicability in dynamic, real-time scenarios. To address this gap, the streaming LLM paradigm has emerged. However, existing definitions of streaming LLMs remain fragmented, conflating streaming generation, streaming inputs, and interactive streaming architectures, while a systematic taxonomy is still lacking. This paper provides a comprehensive overview and analysis of streaming LLMs. First, we establish a unified definition of streaming LLMs based on data flow and dynamic interaction to clarify existing ambiguities. Building on this definition, we propose a systematic taxonomy of current streaming LLMs and conduct an in-depth discussion on their underlying methodologies. Furthermore, we explore the applications of streaming LLMs in real-world scenarios and outline promising research directions to support ongoing advances in streaming intelligence. We maintain a continuously updated repository of relevant papers at https://github.com/EIT-NLP/Awesome-Streaming-LLMs.

Junlong Tong, Zilong Wang, YuJie Ren +4

Shanghai Jiao Tong University · Institute of Digital Twin, Eastern Institute of Technology, Ningbo

Mar 4, 2026

A Multi-Dimensional Quality Scoring Framework for Decentralized LLM Inference with Proof of Quality

Decentralized large language model (LLM) inference networks can pool heterogeneous compute to scale serving, but they require lightweight and incentive-compatible mechanisms to assess output quality. Prior work introduced cost-aware Proof of Quality (PoQ) and adaptive robust PoQ to allocate rewards under evaluator heterogeneity and adversarial behavior. In this paper, we focus on the quality signal itself and propose a multi-dimensional quality scoring framework that decomposes output quality into modular dimensions, including model and cost priors, structure quality, semantic quality, query-output alignment, and agreement/uncertainty. Using logged outputs from QA and summarization tasks, we systematically audit dimension reliability and show that seemingly reasonable dimensions can be task-dependent and even negatively correlated with reference quality without calibration. While the default composite underperforms a strong single semantic evaluator, ablations reveal that removing unreliable dimensions and re-normalizing weights yields a calibrated composite that matches or exceeds the best single- evaluator and consensus baselines. Finally, we integrate the composite score as a drop-in quality signal in PoQ and demonstrate complementary benefits with robust aggregation and adaptive trust weighting under adversarial evaluator attacks.

Arther Tian, Alex Ding, Frank Chen +2

DGrid AI

Language Model Generation Evaluation LLM Evaluation LLM Inference LLM Auditing

Mar 3, 2026

SEALing the Gap: A Reference Framework for LLM Inference Carbon Estimation via Multi-Benchmark Driven Embodiment

Large Language Models are rapidly gaining traction in software engineering, yet their growing carbon footprint raises pressing sustainability concerns. While training emissions are substantial, inference quickly surpasses them due to the sheer volume of prompts processed. This shift underscores the urgent need for accurate, prompt-level carbon measurement during inference to enable informed, sustainability-focused decision-making. To address the limitations of existing approaches, in this paper, we outline the guiding principles for a novel reference framework for LLM inference carbon estimation that can guide the design of future tools and provide a systematic foundation for advancing sustainability research in this domain. We also introduce SEAL, an early embodiment of these principles that leverages a multi-benchmark-driven approach for per-prompt carbon estimation. Its initial validation shows promising results, positioning SEAL as a foundation for standardized sustainability assessment across the LLM ecosystem.

Priyavanshi Pathania, Rohit Mehra, Vibhu Saujanya Sharma +4

Accenture Labs, India · Accenture, USA · Accenture, India

LLM Inference Energy-Efficient ML

Mar 3, 2026

Eliciting Numerical Predictive Distributions of LLMs Without Autoregression

Large Language Models (LLMs) have recently been successfully applied to regression tasks -- such as time series forecasting and tabular prediction -- by leveraging their in-context learning abilities. However, their autoregressive decoding process may be ill-suited to continuous-valued outputs, where obtaining predictive distributions over numerical targets requires repeated sampling, leading to high computational cost and inference time. In this work, we investigate whether distributional properties of LLM predictions can be recovered without explicit autoregressive generation. To this end, we study a set of regression probes trained to predict statistical functionals (e.g., mean, median, quantiles) of the LLM's numerical output distribution directly from its internal representations. Our results suggest that LLM embeddings carry informative signals about summary statistics of their predictive distributions, including the numerical uncertainty. This investigation opens up new questions about how LLMs internally encode uncertainty in numerical tasks, and about the feasibility of lightweight alternatives to sampling-based approaches for uncertainty-aware numerical predictions.

Julianna Piskorz, Katarzyna Kobalczyk, Mihaela van der Schaar

Department of Applied Mathematics and Theoretical Physics, University of Cambridge

LLM Uncertainty Estimation LLM Inference Probabilistic Forecasting Language Model Probing

Mar 2, 2026

Towards Privacy-Preserving LLM Inference via Covariant Obfuscation (Technical Report)

The rapid development of large language models (LLMs) has driven the widespread adoption of cloud-based LLM inference services, while also bringing prominent privacy risks associated with the transmission and processing of private data in remote inference. For privacy-preserving LLM inference technologies to be practically applied in industrial scenarios, three core requirements must be satisfied simultaneously: (1) Accuracy and efficiency losses should be minimized to mitigate degradation in service experience. (2) The inference process can be run on large-scale clusters consist of heterogeneous legacy xPUs. (3) Compatibility with existing LLM infrastructures should be ensured to reuse their engineering optimizations. To the best of our knowledge, none of the existing privacy-preserving LLM inference methods satisfy all the above constraints while delivering meaningful privacy guarantees. In this paper, we propose AloePri, the first privacy-preserving LLM inference method for industrial applications. AloePri protects both the input and output data by covariant obfuscation, which jointly transforms data and model parameters to achieve better accuracy and privacy. We carefully design the transformation for each model component to ensure inference accuracy and data privacy while keeping full compatibility with existing infrastructures of Language Model as a Service. AloePri has been integrated into an industrial system for the evaluation of mainstream LLMs. The evaluation on Deepseek-V3.1-Terminus model (671B parameters) demonstrates that AloePri causes accuracy loss of 0.0%~3.5% and exhibits efficiency equivalent to that of plaintext inference. Meanwhile, AloePri successfully resists state-of-the-art attacks, with less than 5% of tokens recovered. To the best of our knowledge, AloePri is the first method to exhibit practical applicability to large-scale models in real-world systems.

Yu Lin, Qizhi Zhang, Wenqiang Ruan +6

ByteDance · Nanjing University

Privacy-Preserving Language Models LLM Inference Privacy Leakage in Language Models

Mar 1, 2026

Suffix-Constrained Greedy Search Algorithms for Causal Language Models

Large language models (LLMs) are powerful tools that have found applications beyond human-machine interfaces and chatbots. Beside free-form generation, there has been an interest in constrained generation, a setting where LLMs are constrained to generate well-formed outputs with respect to the language defined by a formal grammar. Although appealing, this setting may be over restrictive for downstream applications. For example, many LLM tasks require the model to reason freely before generating its final response in a specific format. In this work, we introduce suffix-constrained generation, a constrained generation setting in which only the end of the response is constrained by a grammar, a scenario that is not supported by existing constrained generation methods. We introduce several suffix-constrained generation algorithms that are based on greedy search. We experiment on several datasets, and show that our approach allows to guarantee suffix constraints without having a negative impact on results, and even improving them in many settings.

Ayoub Hammal, Pierre Zweigenbaum, Caio Corro

1Université Paris-Saclay, CNRS, LISN · 2INSA Rennes, IRISA, CNRS, Université de Rennes

Constrained Generative Modeling Constrained Decoding LLM Inference Controllable Text Generation Autoregressive Decoding

Feb 27, 2026

Architecture-Aware LLM Inference Optimization on AMD Instinct GPUs: A Comprehensive Benchmark and Deployment Study

We present a cross-architecture evaluation of production LLM inference on AMD Instinct MI325X GPUs, benchmarking four models spanning 235B to 1 trillion parameters across three architectural families (MoE+MLA, Dense+GQA, MoE+GQA) on an 8-GPU cluster with 2TB aggregate HBM3e using vLLM v0.14.1. Our results demonstrate that architecture-aware optimization is essential: MLA models require block size 1 and cannot use KV cache offloading, while GQA models benefit from both. The AMD AITER runtime is required for competitive MLA inference throughput and must be selectively disabled for architectures with incompatible attention head configurations. A controlled AITER ablation on Llama-3.1-405B (n=5 per condition) reveals a modest 3-5% throughput benefit at high concurrency but 2-16x higher measurement variability, confirming that AITER's large speedups target MoE/MLA kernels specifically. Under text-only workloads, Llama-405B and DeepSeek V3.2 achieve comparable peak throughput (15,944 and 15,343 tok/s) despite an order-of-magnitude difference in active parameters. Under vision workloads, Qwen3-VL-235B reaches 47,873 tok/s, 6.5x higher than Kimi-K2.5 (7,327 tok/s). Active parameter count per token is associated with inference throughput, though confounded by differences in quantization, AITER acceleration, and tensor parallelism. All four models exhibit a common throughput saturation point consistent with a memory-bandwidth bottleneck (~500 concurrent for short sequences, ~100-200 for longer sequences). All models maintain 100% HTTP-level success rates through 1,000 concurrent users, processing 18.9 million tokens across 17,406 requests without failures.

Athos Georgiou

LLM Inference Efficiency LLM Inference LLM Inference Acceleration AI Accelerator Inference

Feb 27, 2026

PREBA: Surgical Duration Prediction via PCA-Weighted Retrieval-Augmented LLMs and Bayesian Averaging Aggregation

Accurate prediction of surgical duration is pivotal for hospital resource management. Although recent supervised learning approaches-from machine learning (ML) to fine-tuned large language models (LLMs)-have shown strong performance, they remain constrained by the need for high-quality labeled data and computationally intensive training. In contrast, zero-shot LLM inference offers a promising training-free alternative but it lacks grounding in institution-specific clinical context (e.g., local demographics and case-mix distributions), making its predictions clinically misaligned and prone to instability. To address these limitations, we present PREBA, a retrieval-augmented framework that integrates PCA-weighted retrieval and Bayesian averaging aggregation to ground LLM predictions in institution-specific clinical evidence and statistical priors. The core of PREBA is to construct an evidence-based prompt for the LLM, comprising (1) the most clinically similar historical surgical cases and (2) clinical statistical priors. To achieve this, PREBA first encodes heterogeneous clinical features into a unified representation space enabling systematic retrieval. It then performs PCA-weighted retrieval to identify clinically relevant historical cases, which form the evidence context supplied to the LLM. Finally, PREBA applies Bayesian averaging to fuse multi-round LLM predictions with population-level statistical priors, yielding calibrated and clinically plausible duration estimates. We evaluate PREBA on two real-world clinical datasets using three state-of-the-art LLMs, including Qwen3, DeepSeek-R1, and HuatuoGPT-o1. PREBA significantly improves performance-for instance, reducing MAE by up to 40% and raising R^2 from -0.13 to 0.62 over zero-shot inference-and it achieves accuracy competitive with supervised ML methods, demonstrating strong effectiveness and generalization.

Wanyin Wu, Kanxue Li, Baosheng Yu +4

Wanyin Wu, Haoyun Zhao, and Dapeng Tao are with the School of Information Science and Engineering, Yunnan University, Kunming 650500, China. · Kanxue Li and Yibing Zhan are with the School of Computer Science, Wuhan University, Wuhan 430072, China. · Baosheng Yu is with the Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore 639798. +1

Bayesian Model Averaging Clinical Prediction Retrieval-Augmented Generation LLM Inference

Feb 27, 2026

SLA-Aware Distributed LLM Inference Across Device-RAN-Cloud

Embodied AI requires sub-second inference near the Radio Access Network (RAN), but deployments span heterogeneous tiers (on-device, RAN-edge, cloud) and must not disrupt real-time baseband processing. We report measurements from a 5G Standalone (SA) AI-RAN testbed using a fixed baseline policy for repeatability. The setup includes an on-device tier, a three-node RAN-edge cluster co-hosting a containerized 5G RAN, and a cloud tier. We find that on-device execution remains multi-second and fails to meet sub-second budgets. At the RAN edge, SLA feasibility is primarily determined by model variant choice: quantized models concentrate below 0.5,s, while unquantized and some larger quantized models incur deadline misses due to stalls and queuing. In the cloud tier, meeting a 0.5,s deadline is challenging on the measured WAN path (up to 32.9% of requests complete within 0.5,s), but all evaluated variants meet a 1.0,s deadline (100% within 1.0,s). Under saturated downlink traffic and up to

N{=}20

concurrent inference clients, Multi-Instance GPU (MIG) isolation preserves baseband timing-health proxies, supporting safe co-location under fixed partitioning.

Hariz Yet, Nguyen Thanh Tam, Mao V. Ngo +5

Singapore University of Technology and Design (SUTD)

LLM Inference Serving LLM Inference Edge AI Edge-Cloud Computing LLM Inference Scheduling