VibeServe demonstrates that AI agents can synthesize bespoke LLM serving systems end-to-end, remaining competitive with vLLM in standard settings while outperforming it in six non-standard scenarios involving unusual models, workloads, or hardware.
hub Canonical reference
SGLang: Efficient Execution of Structured Language Model Programs
Canonical reference. 100% of citing Pith papers cite this work as background.
81
Pith papers citing it
10
external citations · Pith
Background
100% of classified citations
abstract
Large language models (LLMs) are increasingly used for complex tasks that require multiple generation calls, advanced prompting techniques, control flow, and structured inputs/outputs. However, efficient systems are lacking for programming and executing these applications. We introduce SGLang, a system for efficient execution of complex language model programs. SGLang consists of a frontend language and a runtime. The frontend simplifies programming with primitives for generation and parallelism control. The runtime accelerates execution with novel optimizations like RadixAttention for KV cache reuse and compressed finite state machines for faster structured output decoding. Experiments show that SGLang achieves up to 6.4x higher throughput compared to state-of-the-art inference systems on various large language and multi-modal models on tasks including agent control, logical reasoning, few-shot learning benchmarks, JSON decoding, retrieval-augmented generation pipelines, and multi-turn chat. The code is publicly available at https://github.com/sgl-project/sglang
hub tools
citation-role summary
background 7
citation-polarity summary
roles
background 7polarities
background 7representative citing papers
Knowledge Packs deliver knowledge via pre-computed KV caches with exact equivalence under causal masking, achieving zero divergences on tested questions and enabling value-based steering without training.
Cornfigurator is the first automated deployment planner for generic any-to-any multimodal models that explores the full range of colocation-to-disaggregation strategies and delivers 1.12x to 6.32x higher goodput than existing systems or expert plans.
ELDR reduces median TPOT by 5.9-13.9% in PD-disaggregated MoE serving via expert signatures from prefill, K-means partitioning, and locality-band routing with KV-co-indexed signature cache.
Under GPU-CC, LLM serving losses come from a serialized VM–GPU bridge, not compute; simple scheduling and loader changes recover most of the gap on Blackwell.
Execution-state capsules enable graph-bound full-state checkpointing and sub-millisecond restore for LLMs including KV and recurrent states, yielding 3.9x-27x TTFT speedups in on-device physical-AI serving.
A GEMM-centric taxonomy and unified benchmark show static depth pruning as the strongest Pareto-optimal baseline for LLM inference acceleration, with the frontier shifting to dynamic depth then static width pruning as quality loss rises.
Tangram makes non-uniform KV cache compression practical for LLM serving with deterministic budget allocation, head group paging, and ahead-of-time load balancing, achieving up to 2.6x throughput gains.
QCFuse achieves full-prefill quality in RAG with 1.7x average prefill speedup over full prefill and 1.5x over ProphetKV via compressed query-aware cache fusion.
SiDP distributes model weights across a DP group with WaS and CaS modes to increase KV cache capacity by up to 1.8x and end-to-end throughput by up to 1.5x over vLLM on H20/H200/B200 GPUs for offline LLM inference.
AVMP separates KV and SSM cache pools behind unified virtual addressing with failure-triggered migration, cutting OOM events 7.6% and raising throughput 1.83-13.3x on synthetic loads and 2.36x on ShareGPT traces.
NCCLZ decouples quantization and entropy coding across NCCL stack layers to enable overlapped compression, delivering up to 9.65x speedup over plain NCCL on scientific and training workloads.
EEP makes wide expert-parallel MoE serving survive single-rank failures with an 11s recovery pause, 8s reintegration pause, and throughput restored to 95% of pre-fault level within 52s while staying within 4.4% of a fixed-membership baseline in steady state.
Hosted open-weight LLM APIs function as time-varying heterogeneous services rather than fixed model artifacts, with demand concentrated, supply-use mismatches, and task-specific routing yielding major cost and throughput gains.
Sparse prefix caching via dynamic programming for optimal checkpoint placement under overlap distributions improves the Pareto frontier for recurrent and hybrid LLM serving on shared-prefix data.
TrigReason matches large reasoning model accuracy on math and science benchmarks by delegating most steps to small models and intervening selectively on three triggers, cutting latency by 43.9% and cost by 73.3%.
Fleet adds a Chiplet-task level to GPU task models, enabling per-chiplet scheduling and cooperative cache reuse in persistent megakernels, yielding 1.3-1.5x lower LLM decode latency and up to 37% less HBM traffic on AMD MI350 hardware.
CodeComp uses Joern-extracted Code Property Graph priors for training-free structural KV cache compression, outperforming attention-only baselines on bug localization and code generation while matching full-context patch quality.
SnapStream deploys sparse KV attention in a production inference system on dataflow accelerators, delivering 4x on-chip memory savings for DeepSeek-671B at 128k context with up to 1832 tokens/sec and minimal accuracy loss on LongBench-v2, AIME24, and LiveCodeBench.
MIST is a new simulator for heterogeneous multi-stage LLM inference that combines hardware traces with analytical models to explore configuration trade-offs in hybrid CPU-accelerator systems.
Balancing only first-turn agent requests while keeping follow-ups cache-aware raises LLM cluster TPS 10-16% (colocation) and prefill TPS 2-34% (disaggregation) over SOTA schedulers on production agent traces.
KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.
OmniPilot combines conformal quantile regression with OOD detection to rank LLM serving configurations on mixed GPUs, reporting 6.2% MAPE throughput prediction and 95% top-1 accuracy on 460 benchmark runs while abstaining on unsupported cases.
Speculative pre-positioning decodes stateful sessions ahead with the target model to enable near-constant-time responses from cached distributions or pre-paid deltas at 87% precision for capable models.
citing papers explorer
-
VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?
VibeServe demonstrates that AI agents can synthesize bespoke LLM serving systems end-to-end, remaining competitive with vLLM in standard settings while outperforming it in six non-standard scenarios involving unusual models, workloads, or hardware.
-
Knowledge Packs: Zero-Token Knowledge Delivery via KV Cache Injection
Knowledge Packs deliver knowledge via pre-computed KV caches with exact equivalence under causal masking, achieving zero divergences on tested questions and enabling value-based steering without training.
-
Cornfigurator: Automated Planning for Any-to-Any Multimodal Model Serving
Cornfigurator is the first automated deployment planner for generic any-to-any multimodal models that explores the full range of colocation-to-disaggregation strategies and delivers 1.12x to 6.32x higher goodput than existing systems or expert plans.
-
ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
ELDR reduces median TPOT by 5.9-13.9% in PD-disaggregated MoE serving via expert signatures from prefill, K-means partitioning, and locality-band routing with KV-co-indexed signature cache.
-
The Serialized Bridge: Understanding and Recovering LLM Serving Performance under Blackwell GPU Confidential Computing
Under GPU-CC, LLM serving losses come from a serialized VM–GPU bridge, not compute; simple scheduling and loader changes recover most of the gap on Blackwell.
-
Execution-State Capsules: Graph-Bound Execution-State Checkpoint and Restore for Low-Latency, Small-Batch, On-Device Physical-AI Serving
Execution-state capsules enable graph-bound full-state checkpointing and sub-millisecond restore for LLMs including KV and recurrent states, yielding 3.9x-27x TTFT speedups in on-device physical-AI serving.
-
Beyond FLOPs: Benchmarking Real Inference Acceleration of LLM Pruning under a GEMM-Centric Taxonomy
A GEMM-centric taxonomy and unified benchmark show static depth pruning as the strongest Pareto-optimal baseline for LLM inference acceleration, with the frontier shifting to dynamic depth then static width pruning as quality loss rises.
-
Tangram: Unlocking Non-Uniform KV Cache for Efficient Multi-turn LLM Serving
Tangram makes non-uniform KV cache compression practical for LLM serving with deterministic budget allocation, head group paging, and ahead-of-time load balancing, achieving up to 2.6x throughput gains.
-
QCFuse: Query-Aware Cache Fusion via Compressed View for Efficient RAG Serving
QCFuse achieves full-prefill quality in RAG with 1.7x average prefill speedup over full prefill and 1.5x over ProphetKV via compressed query-aware cache fusion.
-
SiDP: Memory-Efficient Data Parallelism for Offline LLM Inference
SiDP distributes model weights across a DP group with WaS and CaS modes to increase KV cache capacity by up to 1.8x and end-to-end throughput by up to 1.5x over vLLM on H20/H200/B200 GPUs for offline LLM inference.
-
Asymmetric Virtual Memory Paging for Hybrid Mamba-Transformer Inference
AVMP separates KV and SSM cache pools behind unified virtual addressing with failure-triggered migration, cutting OOM events 7.6% and raising throughput 1.83-13.3x on synthetic loads and 2.36x on ShareGPT traces.
-
NCCLZ: Compression-Enabled GPU Collectives with Decoupled Quantization and Entropy Coding
NCCLZ decouples quantization and entropy coding across NCCL stack layers to enable overlapped compression, delivering up to 9.65x speedup over plain NCCL on scientific and training workloads.
-
Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference
EEP makes wide expert-parallel MoE serving survive single-rank failures with an 11s recovery pause, 8s reintegration pause, and throughput restored to 95% of pre-fault level within 52s while staying within 4.4% of a fixed-membership baseline in steady state.
-
When Is the Same Model Not the Same Service? A Measurement Study of Hosted Open-Weight LLM APIs
Hosted open-weight LLM APIs function as time-varying heterogeneous services rather than fixed model artifacts, with demand concentrated, supply-use mismatches, and task-specific routing yielding major cost and throughput gains.
-
Sparse Prefix Caching for Hybrid and Recurrent LLM Serving
Sparse prefix caching via dynamic programming for optimal checkpoint placement under overlap distributions improves the Pareto frontier for recurrent and hybrid LLM serving on shared-prefix data.
-
TrigReason: Trigger-Based Collaboration between Small and Large Reasoning Models
TrigReason matches large reasoning model accuracy on math and science benchmarks by delegating most steps to small models and intervening selectively on three triggers, cutting latency by 43.9% and cost by 73.3%.
-
Fleet: Hierarchical Task-based Abstraction for Megakernels on Multi-Die GPUs
Fleet adds a Chiplet-task level to GPU task models, enabling per-chiplet scheduling and cooperative cache reuse in persistent megakernels, yielding 1.3-1.5x lower LLM decode latency and up to 37% less HBM traffic on AMD MI350 hardware.
-
CodeComp: Structural KV Cache Compression for Agentic Coding
CodeComp uses Joern-extracted Code Property Graph priors for training-free structural KV cache compression, outperforming attention-only baselines on bug localization and code generation while matching full-context patch quality.
-
SnapStream: Efficient Long Sequence Decoding on Dataflow Accelerators
SnapStream deploys sparse KV attention in a production inference system on dataflow accelerators, delivering 4x on-chip memory savings for DeepSeek-671B at 128k context with up to 1832 tokens/sec and minimal accuracy loss on LongBench-v2, AIME24, and LiveCodeBench.
-
MIST: A Co-Design Framework for Heterogeneous, Multi-Stage LLM Inference
MIST is a new simulator for heterogeneous multi-stage LLM inference that combines hardware traces with analytical models to explore configuration trade-offs in hybrid CPU-accelerator systems.
-
SMetric: Rethink LLM Scheduling for Serving Agents with Balanced Session-centric Scheduling
Balancing only first-turn agent requests while keeping follow-ups cache-aware raises LLM cluster TPS 10-16% (colocation) and prefill TPS 2-34% (disaggregation) over SOTA schedulers on production agent traces.
-
What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents
KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.
-
OmniPilot: An Uncertainty-Aware LLM Inference Advisor for Heterogeneous GPU Clusters
OmniPilot combines conformal quantile regression with OOD detection to rank LLM serving configurations on mixed GPUs, reporting 6.2% MAPE throughput prediction and 95% top-1 accuracy on 460 benchmark runs while abstaining on unsupported cases.
-
Speculative Pre-Positioning: Decoding Stateful Sessions to the Next Decision Point Off the Critical Path
Speculative pre-positioning decodes stateful sessions ahead with the target model to enable near-constant-time responses from cached distributions or pre-paid deltas at 87% precision for capable models.
-
KernelSight-LM: A Kernel-Level LLM Inference Simulator
KernelSight-LM simulates LLM inference at kernel granularity with cross-generation (12.1% per-kernel error) and target-measured (3.8% error) tiers, yielding end-to-end median errors of 15.4%/12.8%/3.0% and 14.3%/6.2%/2.7% for TTFT/TPOT/throughput across six model families.
-
FlexMoE: One-for-All Nested Intra-Expert Pruning for MoE Language Models
FlexMoE produces nested pruned subnetworks for MoE LLMs across budgets via channel importance ranking and discrete action learning, plus one mid-budget recovery fine-tune, retaining 99.8% performance at 50% expert parameter pruning.
-
Answer Engineering: Local Trajectory Editing for Protocol-Constrained Decision Making in Large Language Models
Answer Engineering uses local trajectory editing during autoregressive generation to raise protocol compliance on a clinical SSNHL benchmark from 25.1% to 83.5% and balanced accuracy from 42.0% to 80.7%.
-
RKSC: Reasoning-Aware KV Cache Sharing and Confident Early Exit for Multi-Step LLM Inference
RKSC delivers 3.008x mean speedup over baseline and 1.66x over vLLM prefix caching for multi-branch LLM reasoning via similarity-based KV sharing and confidence-gated early exit, with 0.37% error rate.
-
Harnessing Streaming Video in the Wild
Presents Streaming-Train-248K dataset, Streaming Harness system, and Streaming-Eval benchmark to enable VLMs for proactive, memory-equipped streaming video understanding.
-
Sparrow: Sparse Rollout for Stable and Efficient Long-context RL of Large Language Models
Sparrow uses a dynamic sparsity schedule keyed to the lower tail of sparse-to-dense actor-policy mismatch to enable stable and faster rollouts in long-context RL for LLMs.
-
Vortex: Efficient and Programmable Sparse Attention Serving for AI Agents
Vortex provides a programmable frontend and backend for sparse attention in LLM serving, delivering up to 3.46x throughput over full attention while preserving accuracy.
-
MusaCoder: Native GPU Kernel Generation with Full-Stack Training on Moore Threads GPU
MusaCoder combines kernel-oriented data synthesis, diversity-preserving fine-tuning, and stabilized RL with MooreEval to produce correct, fast GPU kernels, with its 27B model setting new SOTA on KernelBench and a MUSA variant.
-
Scaling LLM Inference Beyond Amdahl`s Limits via Eliminating Non-Scalable Overheads
Albireo overlaps non-scalable overheads with compute in tensor-parallel LLM inference to raise the empirical optimal TP degree, delivering up to 1.9x throughput and 48% lower latency versus vLLM.
-
Schedule-Level Shared-Prefix Reuse for LLM RL Training
Schedule-level shared-prefix reuse decouples prefix and suffix passes in GRPO training to compute shared prefixes once, delivering up to 4.395x speedup and 59.1% HBM reduction while preserving numerical equivalence.
-
Draft-OPD: On-Policy Distillation for Speculative Draft Models
Draft-OPD applies on-policy distillation via target-assisted generation and error replay to train speculative draft models, yielding over 5x lossless acceleration and gains over EAGLE-3 and DFlash.
-
A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
-
REVERSE: Reinforcing Evidence Verification and Search for Agentic Image geo-localization
REVERSE uses tool-grounded trajectories and process rewards on visual grounding, query utility, and evidence discrimination to train a 4B model that outperforms retrieval-augmented baselines on Im2GPS3k and YFCC4k.
-
Identifying and Mitigating Systemic Measurement Bias in Production LLM Inference Benchmarks
Standard LLM inference benchmarks introduce systematic bias via GIL-induced queuing in single-process asyncio setups; a multi-process framework and NTPOT metric isolate true serving engine performance at high query rates.
-
Format-Constraint Coupling in Knowledge Graph Construction from Statistical Tables
Empirical 2x2 factorial study on 6 statistical datasets shows format and schema constraints in LLM-based KG construction from CSV tables produce super-additive fidelity loss up to +1.180, with mismatched pairs falling below baseline, plus release of CSVFidelity-Bench.
-
The Constraint Tax: Measuring Validity-Correctness Tradeoffs in Structured Outputs for Small Language Models
Enforcing hard schemas on sub-3B models raises schema validity to 100% but drops answer accuracy from 19.7% to 11.0% and executable accuracy from 91.5% to 48.0% on tool-call tasks.
-
OpenJarvis: Personal AI, On Personal Devices
OpenJarvis decomposes personal AI into Intelligence, Engine, Agents, Tools & Memory, and Learning primitives and applies LLM-guided spec search to produce on-device configurations that reach within 3.2 pp of cloud baselines on average across eight tasks.
-
Attention Once Is All You Need: Efficient Streaming Inference with Stateful Transformers
Stateful sessions with incremental KV cache and flash queries allow O(|q|) latency in streaming transformer inference, delivering up to 5.9x speedup over conventional engines while preserving full attention.
-
Dooly: Configuration-Agnostic, Redundancy-Aware Profiling for LLM Inference Simulation
Dooly reduces LLM inference profiling GPU-hours by 56.4% across 12 models while keeping simulation MAPE under 5% for TTFT and 8% for TPOT by making profiling configuration-agnostic and redundancy-aware.
-
VisMMOE: Exploiting Visual-Expert Affinity for Efficient Visual-Language MoE Offloading
VisMMoE exploits visual-expert affinity via token pruning to achieve up to 2.68x faster VL-MoE inference on memory-constrained hardware while keeping accuracy competitive.
-
RaMP: Runtime-Aware Megakernel Polymorphism for Mixture-of-Experts
RaMP uses a hardware-derived performance region analysis and a four-parameter wave cost model to select optimal polymorphic kernel configurations for MoE inference from runtime expert histograms, delivering 1.22x kernel and 1.30x end-to-end speedups with 0.93% mean regret after brief profiling.
-
When Agents Go Quiet: Output Generation Capacity and Format-Cost Separation for LLM Document Synthesis
LLM agents avoid output stalling and reduce generation tokens by 48-72% via deferred template rendering guided by Output Generation Capacity and a Format-Cost Separation Theorem.
-
ProbeLogits: Kernel-Level LLM Inference Primitives for AI-Native Operating Systems
A kernel-level single-forward-pass logit probe classifies agent actions as safe or dangerous with zero learned parameters and matches or beats Llama Guard 3 on external safety benchmarks.
-
Relax: An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale
Relax is a new RL training engine with omni-native design and async execution that delivers up to 2x speedups over baselines like veRL while converging to equivalent reward levels on Qwen3 models.
-
MEMENTO: Teaching LLMs to Manage Their Own Context
MEMENTO trains LLMs to segment reasoning into blocks, generate mementos as dense summaries, and reason forward using only mementos and KV states, cutting peak KV cache by ~2.5x while preserving benchmark accuracy.
-
Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization
Kernel-Smith combines evolutionary search with RL post-training to generate optimized GPU kernels, achieving SOTA speedups on KernelBench that beat Gemini-3.0-pro and Claude-4.6-opus on NVIDIA Triton and generalize to MetaX MACA.