{"total":248,"rows":[{"id":"2510.03215","title":"Cache-to-Cache: Direct Semantic Communication Between Large Language Models","created":"2025-10-03","cited_by_count":61,"area":"inference"},{"id":"2510.04767","title":"ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMs","created":"2025-10-06","cited_by_count":43,"area":"inference"},{"id":"2601.19278","title":"DART: Diffusion-Inspired Speculative Decoding for Fast LLM Inference","created":"2026-01-27","cited_by_count":24,"area":"inference"},{"id":"2510.07081","title":"Accelerating Diffusion LLM Inference via Local Determinism Propagation","created":"2025-10-08","cited_by_count":21,"area":"inference"},{"id":"2510.00294","title":"Free Draft-and-Verification: Toward Lossless Parallel Decoding for Diffusion Large Language Models","created":"2025-09-30","cited_by_count":19,"area":"inference"},{"id":"2512.03324","title":"Cache What Lasts: Token Retention for Memory-Bounded KV Cache in LLMs","created":"2025-12-03","cited_by_count":19,"area":"inference"},{"id":"2602.08024","title":"FlashVID: Efficient Video Large Language Models via Training-free Tree-based Spatiotemporal Token Merging","created":"2026-02-08","cited_by_count":18,"area":"inference"},{"id":"2602.23881","title":"LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding","created":"2026-02-27","cited_by_count":18,"area":"inference"},{"id":"2605.18810","title":"D-PACE: Dynamic Position-Aware Cross-Entropy for Parallel Speculative Drafting","created":"2026-05-12","cited_by_count":17,"area":"inference"},{"id":"2601.17668","title":"Fast KVzip: Efficient and Accurate LLM Inference with Gated KV Eviction","created":"2026-01-25","cited_by_count":16,"area":"inference"},{"id":"2606.18394","title":"JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting","created":"2026-06-16","cited_by_count":15,"area":"inference"},{"id":"2608.03893","title":"Cross-Model KV Cache Transfer in LLM Families: A Closed-Form Linear Mapping for Prefill Reuse","created":"2026-08-04","cited_by_count":15,"area":"inference"},{"id":"2507.06608","title":"Nexus:Proactive Intra-GPU Disaggregation of Prefill and Decode in LLM Serving","created":"2025-07-09","cited_by_count":13,"area":"inference"},{"id":"2509.18362","title":"FastMTP: Accelerating LLM Inference with Enhanced Multi-Token Prediction","created":"2025-09-16","cited_by_count":13,"area":"inference"},{"id":"2601.04359","title":"PackCache: A Training-Free Acceleration Method for Unified Autoregressive Video Generation via Compact KV-Cache","created":"2026-01-07","cited_by_count":13,"area":"inference"},{"id":"2509.26182","title":"Parallax: Efficient LLM Inference Service over Decentralized Environment","created":"2025-09-30","cited_by_count":12,"area":"inference"},{"id":"2604.09603","title":"ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios","created":"2026-03-10","cited_by_count":12,"area":"inference"},{"id":"2508.08895","title":"ASPD: Unlocking Adaptive Serial-Parallel Decoding by Exploring Intrinsic Parallelism in LLMs","created":"2025-08-12","cited_by_count":11,"area":"inference"},{"id":"2511.20172","title":"Beluga: A CXL-Based Memory Architecture for Scalable and Efficient LLM KVCache Management","created":"2025-11-25","cited_by_count":11,"area":"inference"},{"id":"2604.15039","title":"Prefill-as-a-Service: KVCache of Next-Generation Models Could Go Cross-Datacenter","created":"2026-04-16","cited_by_count":11,"area":"inference"},{"id":"2605.29343","title":"Draft-OPD: On-Policy Distillation for Speculative Draft Models","created":"2026-05-28","cited_by_count":11,"area":"inference"},{"id":"2507.01957","title":"Locality-aware Parallel Decoding for Efficient Autoregressive Image Generation","created":"2025-07-02","cited_by_count":10,"area":"inference"},{"id":"2508.15601","title":"LMDeploy Accelerates Mixed-Precision LLM Inference with TurboMind","created":"2025-08-21","cited_by_count":10,"area":"inference"},{"id":"2509.22134","title":"Bridging Draft Policy Misalignment: Group Tree Optimization for Speculative Decoding","created":"2025-09-26","cited_by_count":10,"area":"inference"},{"id":"2604.09557","title":"SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding","created":"2026-02-10","cited_by_count":10,"area":"inference"},{"id":"2508.16134","title":"CommonKV: Compressing KV Cache with Cross-layer Parameter Sharing","created":"2025-08-22","cited_by_count":9,"area":"inference"},{"id":"2602.08005","title":"ResidualKV: Residual-Based KV Cache Compression for Efficient Long-Context Inference","created":"2026-02-08","cited_by_count":9,"area":"inference"},{"id":"2604.26779","title":"Accelerating RL Post-Training Rollouts via System-Integrated Speculative Decoding","created":"2026-04-29","cited_by_count":9,"area":"inference"},{"id":"2605.17757","title":"OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization","created":"2026-05-18","cited_by_count":9,"area":"inference"},{"id":"2605.08632","title":"PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding","created":"2026-05-09","cited_by_count":8,"area":"inference"},{"id":"2509.17396","title":"EpiCache: Episodic KV Cache Management for Long-Term Conversation on Resource-Constrained Environments","created":"2025-09-22","cited_by_count":7,"area":"inference"},{"id":"2510.20064","title":"Not-a-Bandit: Provably No-Regret Drafter Selection in Speculative Decoding for LLMs","created":"2025-10-22","cited_by_count":7,"area":"inference"},{"id":"2512.02337","title":"SpecPV: Improving Self-Speculative Decoding for Long-Context Generation via Partial Verification","created":"2025-12-02","cited_by_count":7,"area":"inference"},{"id":"2601.17917","title":"Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding","created":"2026-01-25","cited_by_count":7,"area":"inference"},{"id":"2602.06019","title":"Multi-Token Prediction via Self-Distillation","created":"2026-02-05","cited_by_count":7,"area":"inference"},{"id":"2602.07223","title":"Vegas: Self-Speculative Decoding with Verification-Guided Sparse Attention","created":"2026-02-06","cited_by_count":7,"area":"inference"},{"id":"2605.25475","title":"IndexMem: Learned KV-Cache Eviction with Latent Memory for Long-Context LLM Inference","created":"2026-05-25","cited_by_count":7,"area":"inference"},{"id":"2606.06256","title":"RedKnot: Efficient Long-Context LLM Serving with Head-Aware KV Reuse and SegPagedAttention","created":"2026-06-04","cited_by_count":7,"area":"inference"},{"id":"2607.07847","title":"When Does Continual Learning Require Learning","created":"2026-07-08","cited_by_count":7,"area":"inference"},{"id":"2507.23777","title":"XSpecMesh: Quality-Preserving Auto-Regressive Mesh Generation Acceleration via Multi-Head Speculative Decoding","created":"2025-07-31","cited_by_count":6,"area":"inference"},{"id":"2508.15212","title":"SparK: Query-Aware Unstructured Sparsity with Recoverable KV Cache Channel Pruning","created":"2025-08-21","cited_by_count":6,"area":"inference"},{"id":"2509.02121","title":"Batch Query Processing and Optimization for Agentic Workflows","created":"2025-09-02","cited_by_count":6,"area":"inference"},{"id":"2602.21233","title":"AngelSlim: A more accessible, comprehensive, and efficient toolkit for large model compression","created":"2026-02-07","cited_by_count":6,"area":"inference"},{"id":"2605.20104","title":"Draft Less, Retrieve More: Hybrid Tree Construction for Speculative Decoding","created":"2026-05-19","cited_by_count":6,"area":"inference"},{"id":"2606.01563","title":"MomentKV: Closing the Directional Gap in KV Cache Eviction for Long-Context Inference","created":"2026-06-01","cited_by_count":6,"area":"inference"},{"id":"2606.17034","title":"KVEraser: Learning to Steer KV Cache for Efficient Localized Context Erasing","created":"2026-06-15","cited_by_count":6,"area":"inference"},{"id":"2608.13524","title":"DARTree: Speculative Diffusion Decoding with Autoregressive Draft Trees","created":"2026-08-13","cited_by_count":6,"area":"inference"},{"id":"2506.08373","title":"Draft-based Approximate Inference for LLMs","created":"2025-06-10","cited_by_count":5,"area":"inference"},{"id":"2511.00868","title":"FlexiCache: Leveraging Temporal Stability of Attention Heads for Efficient KV Cache Management","created":"2025-11-02","cited_by_count":5,"area":"inference"},{"id":"2511.11907","title":"KVSwap: Disk-aware KV Cache Offloading for Long-Context On-device Inference","created":"2025-11-14","cited_by_count":5,"area":"inference"},{"id":"2512.12284","title":"V-Rex: Real-Time Streaming Video LLM Acceleration via Dynamic KV Cache Retrieval","created":"2025-12-13","cited_by_count":5,"area":"inference"},{"id":"2602.05191","title":"Double-P: Hierarchical Top-P Sparse Attention for Long-Context LLMs","created":"2026-02-05","cited_by_count":5,"area":"inference"},{"id":"2602.20217","title":"KnapSpec: Self-Speculative Decoding via Adaptive Layer Selection as a Knapsack Problem","created":"2026-02-23","cited_by_count":5,"area":"inference"},{"id":"2605.08234","title":"When Does Value-Aware KV Eviction Help? A Fixed-Contract Diagnostic for Non-Monotone Cache Compression","created":"2026-05-07","cited_by_count":5,"area":"inference"},{"id":"2605.19660","title":"OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyond","created":"2026-05-19","cited_by_count":5,"area":"inference"},{"id":"2608.30135","title":"Verification-Aware Training for Speculative Decoding","created":"2026-08-31","cited_by_count":5,"area":"inference"},{"id":"2510.11292","title":"LouisKV: Efficient KV Cache Retrieval for Long Input-Output Sequences","created":"2025-10-13","cited_by_count":4,"area":"inference"},{"id":"2510.13847","title":"DynaSpec: Context-aware Dynamic Speculative Sampling for Large-Vocabulary Language Models","created":"2025-10-11","cited_by_count":4,"area":"inference"},{"id":"2511.00351","title":"Reject Only Critical Tokens: Pivot-Aware Speculative Decoding","created":"2025-11-01","cited_by_count":4,"area":"inference"},{"id":"2511.06174","title":"LUT-LLM: Efficient Large Language Model Inference with Memory-based Computations on FPGAs","created":"2025-11-09","cited_by_count":4,"area":"inference"},{"id":"2511.09844","title":"Steering Pretrained Drafters during Speculative Decoding","created":"2025-11-13","cited_by_count":4,"area":"inference"},{"id":"2511.16193","title":"Fast LLM Post-training via Decoupled and Fastest-of-N Speculation","created":"2025-11-20","cited_by_count":4,"area":"inference"},{"id":"2602.12029","title":"PrefillShare: A Shared Prefill Module for KV Reuse in Multi-LLM Disaggregated Serving","created":"2026-02-12","cited_by_count":4,"area":"inference"},{"id":"2604.04987","title":"Cactus: Accelerating Auto-Regressive Decoding with Constrained Acceptance Speculative Sampling","created":"2026-04-05","cited_by_count":4,"area":"inference"},{"id":"2604.07622","title":"DIVERSED: Relaxed Speculative Decoding via Dynamic Ensemble Verification","created":"2026-04-08","cited_by_count":4,"area":"inference"},{"id":"2605.08317","title":"RDKV: Rate-Distortion Bit Allocation for Joint Eviction and Quantization of the KV Cache","created":"2026-05-08","cited_by_count":4,"area":"inference"},{"id":"2605.23258","title":"A Simple Plug-in for Improving Eviction-Based KV Cache Compression","created":"2026-05-22","cited_by_count":4,"area":"inference"},{"id":"2606.09659","title":"End-to-End Context Compression at Scale","created":"2026-06-08","cited_by_count":4,"area":"inference"},{"id":"2607.01520","title":"The risk of KV cache compression","created":"2026-07-01","cited_by_count":4,"area":"inference"},{"id":"2607.07409","title":"DeLS-Spec: Decoupled Long-Short Contexts for Parallel Speculative Drafting","created":"2026-07-08","cited_by_count":4,"area":"inference"},{"id":"2607.16673","title":"SpecLA: Efficient Speculative Decoding for Linear-Attention Models","created":"2026-07-18","cited_by_count":4,"area":"inference"},{"id":"2607.19223","title":"AdaFlash: Adaptive Speculative Decoding via On-Policy Distilled Diffusion Drafters","created":"2026-07-21","cited_by_count":4,"area":"inference"},{"id":"2608.20530","title":"LiLiCorr: Lightweight Likelihood Correlation of Parallel Drafts for Speculative Decoding","created":"2026-08-20","cited_by_count":4,"area":"inference"},{"id":"2508.17892","title":"ILRe: Intermediate Layer Retrieval for Context Compression in Causal Language Models","created":"2025-08-25","cited_by_count":3,"area":"inference"},{"id":"2509.06261","title":"FineServe: Precision-Aware KV Slab and Two-Level Scheduling for Heterogeneous Precision LLM Serving","created":"2025-09-08","cited_by_count":3,"area":"inference"},{"id":"2510.01336","title":"HiSpec: Hierarchical Speculative Decoding for LLMs","created":"2025-10-01","cited_by_count":3,"area":"inference"},{"id":"2510.14686","title":"xLLM Technical Report","created":"2025-10-16","cited_by_count":3,"area":"inference"},{"id":"2512.14244","title":"From Context to EDUs: Faithful and Structured Context Compression via Elementary Discourse Unit Decomposition","created":"2025-12-16","cited_by_count":3,"area":"inference"},{"id":"2602.06932","title":"When RL Meets Adaptive Speculative Training: A Unified Training-Serving System","created":"2026-02-06","cited_by_count":3,"area":"inference"},{"id":"2602.16603","title":"FlowPrefill: Decoupling Preemption from Prefill Scheduling Granularity to Mitigate Head-of-Line Blocking in LLM Serving","created":"2026-02-18","cited_by_count":3,"area":"inference"},{"id":"2603.05663","title":"Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal Grounding","created":"2026-03-05","cited_by_count":3,"area":"inference"},{"id":"2605.06675","title":"RateQuant: Optimal Mixed-Precision KV Cache Quantization via Rate-Distortion Theory","created":"2026-04-22","cited_by_count":3,"area":"inference"},{"id":"2605.07243","title":"SpecBlock: Block-Iterative Speculative Decoding with Dynamic Tree Drafting","created":"2026-05-08","cited_by_count":3,"area":"inference"},{"id":"2605.28069","title":"ZipRL: Adaptive Multi-Turn Context Compression with Hindsight Response Replay","created":"2026-05-27","cited_by_count":3,"area":"inference"},{"id":"2607.08690","title":"A Practical Investigation of Training-free Relaxed Speculative Decoding","created":"2026-07-09","cited_by_count":3,"area":"inference"},{"id":"2607.10186","title":"FlashAccel: Leveraging High-Bandwidth Flash (HBF) for High-Throughput LLM Inference","created":"2026-07-11","cited_by_count":3,"area":"inference"},{"id":"2608.01247","title":"RestoreKV: Recovering Full-Cache Behavior Under Aggressive Query-Agnostic KV Cache Eviction","created":"2026-08-02","cited_by_count":3,"area":"inference"},{"id":"2608.02123","title":"From Chains to Trees: Parent-Conditioned Drafting for Semi-Autoregressive Speculative Decoding","created":"2026-08-03","cited_by_count":3,"area":"inference"},{"id":"2508.03611","title":"Astrolabe: Balancing Load in LLM Serving with Randomized Prediction-Guided Scheduling","created":"2025-08-05","cited_by_count":2,"area":"inference"},{"id":"2509.04827","title":"VoltanaLLM: Energy-Efficient and SLO-Aware Disaggregated LLM Serving via Adaptive Frequency Control and State-Space Routing","created":"2025-09-05","cited_by_count":2,"area":"inference"},{"id":"2511.03092","title":"SnapStream: Efficient Long Sequence Decoding on Dataflow Accelerators","created":"2025-11-05","cited_by_count":2,"area":"inference"},{"id":"2512.04857","title":"Autoregressive Image Generation Needs Only a Few Lines of Cached Tokens","created":"2025-12-04","cited_by_count":2,"area":"inference"},{"id":"2512.11529","title":"xGR: Efficient Generative Recommendation Serving at Scale","created":"2025-12-12","cited_by_count":2,"area":"inference"},{"id":"2602.01274","title":"PACER: Blockwise Pre-verification for Speculative Decoding with Adaptive Length","created":"2026-02-01","cited_by_count":2,"area":"inference"},{"id":"2602.07721","title":"ParisKV: Fast and Drift-Robust KV-Cache Retrieval for Long-Context LLMs","created":"2026-02-07","cited_by_count":2,"area":"inference"},{"id":"2602.07878","title":"Rethinking Latency Denial-of-Service: Attacking the LLM Serving Framework, Not the Model","created":"2026-02-08","cited_by_count":2,"area":"inference"},{"id":"2603.02188","title":"Multi-Head Low-Rank Attention","created":"2026-03-02","cited_by_count":2,"area":"inference"},{"id":"2604.01881","title":"HieraVid: Hierarchical Token Pruning for Fast Video Large Language Models","created":"2026-04-02","cited_by_count":2,"area":"inference"},{"id":"2604.02047","title":"Goose: Anisotropic Speculation Trees for Training-Free Speculative Decoding","created":"2026-04-02","cited_by_count":2,"area":"inference"},{"id":"2604.06664","title":"Foundry: Template-Based CUDA Graph Context Materialization for Fast LLM Serving Cold Start","created":"2026-04-08","cited_by_count":2,"area":"inference"}],"next_cursor":"100"}