Skip to yearly menu bar Skip to main content


Show Detail
Schedule Mon Tue Wed Thu Fri
Timezone: US/Pacific
 
Filter Rooms:  

SUN 17 MAY
3 p.m.
Registration Open
(ends 6:00 PM)

MON 18 MAY
8 a.m.
Registration
(ends 5:00 PM)
8:50 a.m.
Remarks:
Opening Remarks
Grand Ballroom 1
(ends 9:00 AM)
9 a.m.
Invited Talk:
Rethinking Open Source Contribution in the Age of AI Agents
Grand Ballroom 1
Roger Wang
(ends 9:25 AM)
9:25 a.m.
Invited Talk:
Beyond Model Serving: Cross-Stack Co-Design for Agentic Systems
Grand Ballroom 1
Esha Choukse
(ends 9:50 AM)
9:50 a.m.
Invited Talk:
LMCache: An Efficient KV Cache Layer for Enterprise-Scale LLM Inference
Grand Ballroom 1
Yuhan Liu
(ends 10:15 AM)
10:15 a.m.
Invited Talk:
Eliciting Language Model Behaviors with Investigator Agents
Grand Ballroom 1
Lisa Li
(ends 10:40 AM)
10:40 a.m.
Break:
Coffee Break
(ends 11:00 AM)
11 a.m.
noon
Break:
Lunch (on your own)
(ends 1:30 PM)
1:30 p.m.
Keynote Talk:
When AI Starts Writing Systems Code
Grand Ballroom 1
Mark Saroufim
(ends 2:30 PM)
2:30 p.m.
3:35 p.m.
Break:
Coffee Break
(ends 4:00 PM)
4 p.m.
Panel:
Panel
Grand Ballroom 1
(ends 5:00 PM)
5 p.m.
YPS Poster Session & YPS Reception [5:00-7:00]
Evergreen Ballroom
Posters -
Impact of Scheduling for Terminal Agent Workloads on Unified-Memory Workstations
Tiered Autonomy Framework for Human–Agent Collaboration in Mission-Critical Cyber-Physical Systems
On the Diminishing Returns of Expert Load Balancing in MoE LLM Serving
Neuro-Analog
SAT-Eval: A Framework for Preference Drift in Multi-Turn LLM Conversations
Learning-Guided Design Optimization for Lifecycle Impact Analysis
Cascade: Utility-Driven Speculative Decoding for Mixture-of-Experts
LYNX: Workload-Agnostic Expert Remapping for Efficient MoE Inference
Automated Feature Engineering -- Faster Iteration to Solve Business Problems
ADAPTIVE ERASURE CODING FOR FAULT-TOLERANT LLM SERVING WITH CONTINUOUS BATCHING
Practical Unstructured Sparsity for Efficient LLM Inference
HiSpec: Hierarchical Speculative Decoding for LLMs
HiServe: A Prefix Cache Serving System for Hybrid LLMs
BioTriton: Portable Cross-Vendor GPU Kernels for High-Throughput Bioinformatics via OpenAI Triton
REMIX: Dynamic Partitioning for Fine-Grained Heterogeneous LLM Serving
Toward a Small ML Runtime Stack for Raspberry Pi 5 QPUs
Communication-Efficient Distributed Inference for Transformer Models via Vector Quantized Context
LearnedCache: An eBPF-Integrated Perceptron-Based Eviction Policy for the Linux Page Cache
Accelerating LLM Inference: Self-Speculative Decoding via Learned Seed Injection
HADIS: Hybrid Adaptive Diffusion Model Serving for Efficient Text-to-Image Generation
Leveraging ASIC AI Chips for Homomorphic Encryption
Shortcut-connected Expert Parallelism for Accelerating Mixture of Experts
Speciesism in the Assistant Axis: Probing Compassion Vectors in Post-Trained LLMs
ov_training_kit : Model training and inference on local AI PC to strengthen the AI ecosystem
NeSyKV: Neuro-Symbolic Architecture-Specific KV-Cache Eviction for LLM Inference
DriftBench: Measuring and Predicting Infrastructure Drift in LLM Serving Systems
ViRuleEval: A Neuro-Symbolic System for Interpretable Evaluation of Text-to-Video Generation
Kascade: A Practical Sparse Attention Method for Long-Context LLM Inference
Designing Communication-Efficient AI Systems: An Interconnect-Aware HPC Perspective
AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimization
Equinox: Decentralized Scheduling for Hardware-aware Satellite Intelligence
BLAZE: Bias-Driven Load-Aware Zero-Overhead Expert Routing
ForeCache: Understanding Workloads and Optimizing KVCache Management for Efficiently Serving LLM Coding Agents
SD-HC: Heterogeneous Functional Pipelining for Speculative LLM Decoding on AI PCs
Flexo: A User-Controllable Distributed Training System
From 805 ms to 23 ms: Accelerating State-Space Models for Real-Time ICU Monitoring with Fused Triton Kernels
A Framework for Evaluating Neural Network Deployability on Analog In-Memory Computing Hardware
Towards Efficient Systems for Long-Context Automatic Speech Recognition
(ends 7:00 PM)

TUE 19 MAY
8 a.m.
Registration
(ends 5:00 PM)
8:30 a.m.
Remarks:
Opening Remarks
Grand Ballroom 1
(ends 8:45 AM)
8:45 a.m.
Research Track Oral Presentation: Best Paper Session [8:45-10:05]
Grand Ballroom 1
Moderator: Dimitrios Skarlatos
StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation
LEANN: A Low-Storage Overhead Vector Index
BLASST: Dynamic BLocked Attention Sparsity via Softmax Thresholding
ExecuTorch - A Unified PyTorch Solution to Run ML Models On-Device
(ends 10:05 AM)
Overflow / Viewing [8:45-10:05]
Grand Ballroom 2
(ends 10:05 AM)
10:05 a.m.
Break:
Coffee Break
(ends 10:30 AM)
10:30 a.m.
Keynote Talk:
The Next Horizon of Systems: From MLSys to System Intelligence
Grand Ballroom 1
Lidong Zhou
(ends 11:30 AM)
Overflow / Viewing [10:30-11:30]
Grand Ballroom 2
(ends 11:30 AM)
11:30 a.m.
Break:
Lunch (on your own)
(ends 1:00 PM)
1 p.m.
Research Track Oral Presentation: Agentic AI 1 [1:00-2:15]
Grand Ballroom 1
Moderator: Cheng Tan
OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents
VeriMoA: A Mixture-of-Agents Framework for Spec-to-HDL Generation
When Enough is Enough: Rank-Aware Early Termination for Vector Search
Optimizing PyTorch Inference with LLM-Based Multi-Agent Systems
Matrix: Peer-to-Peer Multi-Agent Synthetic Data Generation Framework
(ends 2:15 PM)
Research Track Oral Presentation: Multimodal and Generative Models [1:00-2:15]
Grand Ballroom 2
Moderator: Haoran Qiu
ContextPilot: Fast Long-Context Inference via Context Reuse
db-SP: Accelerating Sparse Attention for Visual Generative Models with Dual-Balanced Sequence Parallelism
SwiftGS: Algorithm and System Co-Optimization for Fast 3D Gaussian Splatting on GPUs
TriInfer: Hybrid EPD Disaggregation for Efficient Multimodal Large Language Model Inference
TiDAR: Think in Diffusion, Talk in Autoregression
(ends 2:15 PM)
2:15 p.m.
Break:
Coffee Break
(ends 2:45 PM)
2:45 p.m.
Research Track Oral Presentation: Agentic AI 2 [2:45-4:00]
Grand Ballroom 1
Moderator: Yawen Wang
Hippocampus: An Efficient and Scalable Memory Module for Agentic AI
AgenticCache: Cache-Driven Asynchronous Planning for Embodied AI Agents
TeleRAG: Efficient Retrieval-Augmented Generation Inference with Lookahead Retrieval
FlashAgents: Accelerating Multi-Agent LLM Systems via Streaming Prefill Overlap
Ontology-Guided Long-Term Agent Memory for Conversational RAG
(ends 4:00 PM)
Research Track Oral Presentation: LLM Serving 1 [2:45-4:00]
Grand Ballroom 2
Moderator: Esha Choukse
Stream2LLM: Overlap Context Streaming and Prefill for Reduced Time-to-First-Token
SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips
PLA-Serve: A Prefill-Length-Aware LLM Serving System
TokenWeave: Efficient Compute-Communication Overlap for Distributed LLM Inference
From Tokens to Layers: Redefining Stall-Free Scheduling for MoE Serving with Layered Prefill
(ends 4:00 PM)
4 p.m.
Break:
Break
(ends 4:30 PM)
4:30 p.m.
Research Track Oral Presentation: Security and Privacy [4:30-6:00]
Grand Ballroom 2
Moderator: Runzhou Tao
G-HEMP: FAST MULTI-GPU PRIVATE INFERENCE FOR LARGE-SCALE GCNS WITH HOMOMORPHIC ENCRYPTION
CSLE: A Reinforcement Learning Platform for Autonomous Security Management
Privatar: Scalable Privacy-preserving Multi-user VR via Secure Offloading
ZK-APEX: ZERO-KNOWLEDGE APPROXIMATE PERSONALIZED UNLEARNING WITH EXECUTABLE PROOFS
Blueprint, Bootstrap, and Bridge: A Security Look at NVIDIA GPU Confidential Computing
Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle Problem
(ends 6:00 PM)
Research Track Oral Presentation: Federated Learning [4:30-6:00]
Grand Ballroom 1
Moderator: Hui Guan
SONAR: Benchmarking Topology and Collaboration in Decentralized Learning
ProToken: Token-Level Attribution for Federated Large Language Models
FLoRIST: Singular Value Thresholding for Efficient and Accurate Federated Fine-Tuning of Large Language Models
DisAgg: Distributed Aggregators for Efficient Secure Aggregation
PLayer-FL: A Principled Approach to Personalized Layer-wise Cross-Silo Federated Learning
Zero redundancy distributed learning with differential privacy
(ends 6:00 PM)
6 p.m.
Poster Session 1 & Opening Reception [6:00-8:00]
Evergreen Ballroom
Posters -
StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation
LEANN: A Low-Storage Overhead Vector Index
BLASST: Dynamic BLocked Attention Sparsity via Softmax Thresholding
ExecuTorch - A Unified PyTorch Solution to Run ML Models On-Device
OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents
ContextPilot: Fast Long-Context Inference via Context Reuse
VeriMoA: A Mixture-of-Agents Framework for Spec-to-HDL Generation
db-SP: Accelerating Sparse Attention for Visual Generative Models with Dual-Balanced Sequence Parallelism
When Enough is Enough: Rank-Aware Early Termination for Vector Search
SwiftGS: Algorithm and System Co-Optimization for Fast 3D Gaussian Splatting on GPUs
Optimizing PyTorch Inference with LLM-Based Multi-Agent Systems
TriInfer: Hybrid EPD Disaggregation for Efficient Multimodal Large Language Model Inference
Matrix: Peer-to-Peer Multi-Agent Synthetic Data Generation Framework
TiDAR: Think in Diffusion, Talk in Autoregression
Hippocampus: An Efficient and Scalable Memory Module for Agentic AI
Stream2LLM: Overlap Context Streaming and Prefill for Reduced Time-to-First-Token
SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips
AgenticCache: Cache-Driven Asynchronous Planning for Embodied AI Agents
TeleRAG: Efficient Retrieval-Augmented Generation Inference with Lookahead Retrieval
PLA-Serve: A Prefill-Length-Aware LLM Serving System
FlashAgents: Accelerating Multi-Agent LLM Systems via Streaming Prefill Overlap
TokenWeave: Efficient Compute-Communication Overlap for Distributed LLM Inference
Ontology-Guided Long-Term Agent Memory for Conversational RAG
From Tokens to Layers: Redefining Stall-Free Scheduling for MoE Serving with Layered Prefill
SONAR: Benchmarking Topology and Collaboration in Decentralized Learning
G-HEMP: FAST MULTI-GPU PRIVATE INFERENCE FOR LARGE-SCALE GCNS WITH HOMOMORPHIC ENCRYPTION
ProToken: Token-Level Attribution for Federated Large Language Models
CSLE: A Reinforcement Learning Platform for Autonomous Security Management
FLoRIST: Singular Value Thresholding for Efficient and Accurate Federated Fine-Tuning of Large Language Models
Privatar: Scalable Privacy-preserving Multi-user VR via Secure Offloading
DisAgg: Distributed Aggregators for Efficient Secure Aggregation
ZK-APEX: ZERO-KNOWLEDGE APPROXIMATE PERSONALIZED UNLEARNING WITH EXECUTABLE PROOFS
PLayer-FL: A Principled Approach to Personalized Layer-wise Cross-Silo Federated Learning
Blueprint, Bootstrap, and Bridge: A Security Look at NVIDIA GPU Confidential Computing
Zero redundancy distributed learning with differential privacy
Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle Problem
(ends 8:00 PM)

WED 20 MAY
8 a.m.
Registration
(ends 5:00 PM)
8:30 a.m.
Research Track Oral Presentation: LLM Serving 2 [8:30-10:00]
Grand Ballroom 1
Moderator: Juncheng Yang
BEAM: Joint Resource–Power Optimization for Energy-Efficient LLM Inference under SLO contraints
MorphServe: Efficient and Workload-Aware LLM Serving via Runtime Quantized Layer Swapping and KV Cache Resizing
BOute: Cost-Efficient LLM Serving with Heterogeneous LLMs and GPUs via Multi-Objective Bayesian Optimization
Locality-Aware Beam Scheduling for Efficient Test-Time Compute with a Consumer-grade GPU
Breaking the Ice: Analyzing Cold Start Latency in vLLM
FaaScale: Unlocking Fast LLM Scaling for Serverless Inference
(ends 10:00 AM)
Research Track Oral Presentation: LLM Training 1 [8:30-10:00]
Grand Ballroom 2
Moderator: Cheng Tan
BOOST: BOttleneck-Optimized Scalable Training Framework for Low-Rank Large Language Models
Unleashing Scalable Context Parallelism for Foundation Models Pre-Training via FCP
ProTrain: Efficient LLM Training via Automatic Memory Management
MTraining: Distributed Dynamic Sparse Attention for Efficient Ultra-Long Context Training
NEST: Network- and Memory-Aware Device Placement for Distributed Deep Learning
Efficient Long-Context Language Model Training by Core Attention Disaggregation
(ends 10:00 AM)
10 a.m.
Break:
Coffee Break
(ends 10:30 AM)
10:30 a.m.
Keynote Talk:
Keynote: Amin Vahdat - SVP and Chief Technologist, AI & Infrastructure
Grand Ballroom 1
Amin Vahdat
(ends 11:30 AM)
Overflow / Viewing [10:30-11:30]
Grand Ballroom 2
(ends 11:30 AM)
11:30 a.m.
Fireside Chat:
Fireside Chat
Grand Ballroom 1
(ends 12:15 PM)
12:15 p.m.
Break:
Lunch (on your own)
(ends 1:30 PM)
1:30 p.m.
Research Track Oral Presentation: LLM Serving 3 [1:30-3:00]
Grand Ballroom 1
Moderator: Yifan Qiao
SkipKV: Selective Skipping of KV Generation and Storage for Efficient Inference with Large Reasoning Models
OPKV: A High-Throughput Plugin-Driven Framework for Recallable Sparsity in Paged KV Cache Systems
FlexiCache: Leveraging Temporal Stability of Attention Heads for Efficient KV Cache Management
MAC-Attention: a Match--Amend--Complete scheme for fast and accurate attention computation
Using Span Queries to Optimize Cache and Attention Locality
Kitty: Accurate and Efficient 2-bit KV Cache Quantization with Dynamic Channel-wise Precision Boost
(ends 3:00 PM)
Research Track Oral Presentation: ML for Systems [1:30-3:00]
Grand Ballroom 2
Moderator: Hui Guan
FlashInfer-Bench: Building the Virtuous Cycle for AI-driven LLM Systems
Virtual Machine NUMA Placement at Scale: Learning the Norm, Shielding the Tail
When Machine Learning Isn’t Sure: Building Resilient ML-Based Computer Systems by Embracing Uncertainty
Practical Adversarial Multi-Armed Bandits with Sublinear Runtime
Unified LLM Model for Power, Performance, and Area Prediction from Hardware Code
Automated Algorithm Design for Auto-Tuning Optimizers
(ends 3:00 PM)
3 p.m.
Break:
Coffee Break
(ends 3:15 PM)
3:15 p.m.
Research Track Oral Presentation: LLM Serving 4 [3:15-4:45]
Grand Ballroom 1
Moderator: Christina Giannoula
RaidServe: High-performance Resilient Serving
fabric-lib: RDMA Point-to-Point Communication for LLM Systems
Demystifying the Mixture of Experts Serving Tax
FarSkip-Collective: Unhobbling Blocking Communication in Mixture of Experts Models
GhostServe: A Lightweight Checkpointing System in the Shadow for Fault-Tolerant LLM Serving
CRAFT: Fine-Grained Cost-Aware Expert Replication For Efficient Mixture-of-Experts Serving
(ends 4:45 PM)
Research Track Oral Presentation: LLM Training 2 [3:15-4:45]
Grand Ballroom 2
Moderator: Mohamed Abdelfattah
ReSpec: Towards Optimizing Speculative Decoding in Reinforcement Learning Systems
HetRL: Efficient Reinforcement Learning for LLMs in Heterogeneous Environments
Pylo: Towards Accessible Learned Optimizers in PyTorch
Learning from Less: Measuring the Effectiveness of RLVR in Low Data and Compute Regimes
FlexTrain: Scalable Hybrid-Parallel Training with Elastic Resource Utilization and Consistent Accuracy
Beat the long tail: Distribution-Aware Speculative Decoding for RL Training
(ends 4:45 PM)
4:45 p.m.
Break:
Break
(ends 5:00 PM)
5 p.m.
Research Track Oral Presentation: LLM Training 3 [5:00-6:30]
Grand Ballroom 1
Moderator: Zhihao Zhang
Zorse: Optimizing LLM Training Efficiency on Heterogeneous GPU Clusters
GriNNder: Breaking the Memory Capacity Wall in Full-Graph GNN Training with Storage Offloading
HexiScale: Facilitating Large Language Model Training over Heterogeneous Hardware
A Lightweight High-Throughput Collective-Capable NoC for Large-Scale ML Accelerators
DreamDDP: Accelerating Low-Bandwidth Geo-Distributed LLM Training with Layer-wise Partial Synchronization
FP8-Flow-MoE: A Casting-Free FP8 Recipe without Double Quantization Error
(ends 6:30 PM)
Research Track Oral Presentation: Model Compression [5:00-6:30]
Grand Ballroom 2
Moderator: Sanjeev Singh
CAGE: Curvature-Aware Gradient Estimation For Accurate Quantization-Aware Training
Shannonic: Efficient Entropy-Optimal Compression for ML Workloads
Once-for-All Channel Mixers (HyperTinyPW): Generative Compression for TinyML
MixLLM: LLM Quantization with Global Mixed-precision between Output-features and Highly-efficient System Design
Search Your Block Floating Point Scales!
(ends 6:30 PM)
6:30 p.m.
Poster Session 2 [6:30-8:30]
Evergreen Ballroom
Posters -
BEAM: Joint Resource–Power Optimization for Energy-Efficient LLM Inference under SLO contraints
BOOST: BOttleneck-Optimized Scalable Training Framework for Low-Rank Large Language Models
Unleashing Scalable Context Parallelism for Foundation Models Pre-Training via FCP
MorphServe: Efficient and Workload-Aware LLM Serving via Runtime Quantized Layer Swapping and KV Cache Resizing
ProTrain: Efficient LLM Training via Automatic Memory Management
BOute: Cost-Efficient LLM Serving with Heterogeneous LLMs and GPUs via Multi-Objective Bayesian Optimization
Locality-Aware Beam Scheduling for Efficient Test-Time Compute with a Consumer-grade GPU
MTraining: Distributed Dynamic Sparse Attention for Efficient Ultra-Long Context Training
Breaking the Ice: Analyzing Cold Start Latency in vLLM
NEST: Network- and Memory-Aware Device Placement for Distributed Deep Learning
FaaScale: Unlocking Fast LLM Scaling for Serverless Inference
Efficient Long-Context Language Model Training by Core Attention Disaggregation
SkipKV: Selective Skipping of KV Generation and Storage for Efficient Inference with Large Reasoning Models
FlashInfer-Bench: Building the Virtuous Cycle for AI-driven LLM Systems
OPKV: A High-Throughput Plugin-Driven Framework for Recallable Sparsity in Paged KV Cache Systems
Virtual Machine NUMA Placement at Scale: Learning the Norm, Shielding the Tail
FlexiCache: Leveraging Temporal Stability of Attention Heads for Efficient KV Cache Management
When Machine Learning Isn’t Sure: Building Resilient ML-Based Computer Systems by Embracing Uncertainty
MAC-Attention: a Match--Amend--Complete scheme for fast and accurate attention computation
Practical Adversarial Multi-Armed Bandits with Sublinear Runtime
Unified LLM Model for Power, Performance, and Area Prediction from Hardware Code
Using Span Queries to Optimize Cache and Attention Locality
Automated Algorithm Design for Auto-Tuning Optimizers
Kitty: Accurate and Efficient 2-bit KV Cache Quantization with Dynamic Channel-wise Precision Boost
RaidServe: High-performance Resilient Serving
ReSpec: Towards Optimizing Speculative Decoding in Reinforcement Learning Systems
HetRL: Efficient Reinforcement Learning for LLMs in Heterogeneous Environments
fabric-lib: RDMA Point-to-Point Communication for LLM Systems
Pylo: Towards Accessible Learned Optimizers in PyTorch
Demystifying the Mixture of Experts Serving Tax
Learning from Less: Measuring the Effectiveness of RLVR in Low Data and Compute Regimes
FarSkip-Collective: Unhobbling Blocking Communication in Mixture of Experts Models
FlexTrain: Scalable Hybrid-Parallel Training with Elastic Resource Utilization and Consistent Accuracy
GhostServe: A Lightweight Checkpointing System in the Shadow for Fault-Tolerant LLM Serving
Beat the long tail: Distribution-Aware Speculative Decoding for RL Training
CRAFT: Fine-Grained Cost-Aware Expert Replication For Efficient Mixture-of-Experts Serving
Zorse: Optimizing LLM Training Efficiency on Heterogeneous GPU Clusters
CAGE: Curvature-Aware Gradient Estimation For Accurate Quantization-Aware Training
GriNNder: Breaking the Memory Capacity Wall in Full-Graph GNN Training with Storage Offloading
Shannonic: Efficient Entropy-Optimal Compression for ML Workloads
HexiScale: Facilitating Large Language Model Training over Heterogeneous Hardware
Once-for-All Channel Mixers (HyperTinyPW): Generative Compression for TinyML
MixLLM: LLM Quantization with Global Mixed-precision between Output-features and Highly-efficient System Design
A Lightweight High-Throughput Collective-Capable NoC for Large-Scale ML Accelerators
DreamDDP: Accelerating Low-Bandwidth Geo-Distributed LLM Training with Layer-wise Partial Synchronization
Search Your Block Floating Point Scales!
FP8-Flow-MoE: A Casting-Free FP8 Recipe without Double Quantization Error
(ends 8:30 PM)

THU 21 MAY
8 a.m.
Registration
(ends 5:00 PM)
8:30 a.m.
Industry Track Oral Presentation: Agentic AI/MLSys [8:30-10:00]
Grand Ballroom 1
Moderator: Binhang Yuan
ADR: AN AGENTIC DETECTION SYSTEMFORENTERPRISE AGENTIC AI SECURITY
PROMPTS: PeRformance Optimization via Multi-Agent Planning for LLM Training and Serving
Agentic Operator Generation for ML ASICs
Cost-aware Duration Prediction for Software Upgrades in Datacenters
The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents
BatchLLM: Optimizing Large Batched LLM Inference with Global Prefix Sharing and Throughput-oriented Token Batching
(ends 10:00 AM)
Research Track Oral Presentation: LLM Serving 5 [8:30-10:00]
Grand Ballroom 2
Moderator: Xiang Song
PRISM: Parametrically Refactor Inference for Speculative Decoding Draft Models
CDLM: Consistency Diffusion Language Models for Faster Sampling
Speculative Decoding: Performance or Illusion?
SpecDiff-2: Scaling Diffusion Drafter Alignment For Faster Speculative Decoding
Accelerating Large-Scale Reasoning Model Inference with Sparse Self-Speculative Decoding
(ends 10:00 AM)
10 a.m.
Break:
Coffee Break
(ends 10:30 AM)
10:30 a.m.
Keynote Talk:
Rethinking Pretraining: Data and Architecture
Grand Ballroom 1
Luke Zettlemoyer
(ends 11:30 AM)
Overflow / Viewing [10:30-11:30]
Grand Ballroom 2
(ends 11:30 AM)
11:30 a.m.
Break:
Lunch (on your own)
(ends 1:00 PM)
1 p.m.
Industry Track Oral Presentation: LLM Serving 6 [1:00-2:15]
Grand Ballroom 1
Moderator: Esha Choukse
SHIP: SRAM-Based Huge Inference Pipelines for Fast LLM Serving
Beyond the Buzz: A Pragmatic Take on Inference Disaggregation
Optimizing Deployment Configurations for LLM Inference
Meeting SLOs, Slashing Hours: Automated Enterprise LLM Optimization with OptiKIT
Scaling Up Large Language Models Serving Systems for Semantic Job Search
(ends 2:15 PM)
Research Track Oral Presentation: Efficient ML [1:00-2:15]
Grand Ballroom 2
Moderator: Vijay Janapa Reddi
HELIOS : Adaptive Model And Early-Exit Selection for Efficient LLM Inference Serving
IntAttention: A Fully Integer Attention Pipeline for Efficient Edge Inference
REPARO: LOSS-RESILIENT GENERATIVE CODEC FOR VIDEO CONFERENCING
Rethinking DVFS for Mobile LLMs: Unified Energy-Aware Scheduling with CORE
EarthSight: A Distributed Framework for Low-Latency Satellite Intelligence
(ends 2:15 PM)
2:15 p.m.
Break:
Coffee Break
(ends 2:45 PM)
2:45 p.m.
Industry Track Oral Presentation: LLM Training 4 [2:45-4:15]
Grand Ballroom 1
Moderator: Yanghua Peng
Sparing Strategies to Minimize Reliability Impact On Large Training Jobs
veScale-FSDP: Flexible and High-Performance FSDP at Scale
AXLearn: Modular, Hardware-Agnostic Large Model Training
GUARD: SCALABLE STRAGGLER DETECTION AND NODE HEALTH MANAGEMENT FOR LARGE-SCALE TRAINING
MoEBlaze: Breaking the Memory Wall for Efficient MoE Training on Modern GPUs
FreeScale: Distributed Training for Sequence Recommendation Models with Minimal Scaling Cost
(ends 4:15 PM)
Research Track Oral Presentation: Benchmarks and Evaluation [2:45-4:15]
Grand Ballroom 2
Moderator: Juncheng Yang
Massive-Scale Out-Of-Core UMAP on the GPU
Charon: A Unified and Fine-Grained Simulator for Large-Scale LLM Training and Inference
Hawkeye: Reproducing GPU-Level Non-Determinism
DriftBench: Measuring and Predicting Infrastructure Drift in LLM Serving Systems
PARROT: Persuasion and Agreement Robustness Rating of Output Truth — A Sycophancy Robustness Benchmark for LLMs
(ends 4:15 PM)
4:15 p.m.
Break:
Break
(ends 4:30 PM)
4:30 p.m.
Industry Track Oral Presentation: Benchmarks and Evaluation [4:30-6:00]
Grand Ballroom 1
Moderator: Youjie Li
XProf: An Open, Scalable, and Extensible Profiling System for the Modern ML Stack
AIRS: Scaling Live Inference in Resource Constrained Environments
SAKURAONE: An Open Ethernet–Based AI HPC System and Its Observed Workload Dynamics in a Single-Tenant LLM Development Environment
MLCommons Chakra: Advancing Performance Benchmarking and Co-design using Standardized Execution Traces
ProfInfer: An eBPF-based Fine-Grained LLM Inference Profiler
Machine Learning Fleet Efficiency: Improving TPU Systems at Scale with ML Productivity Goodput
(ends 6:00 PM)
Research Track Oral Presentation: Compilers and Kernels [4:30-6:00]
Grand Ballroom 2
Moderator: Steven Lyubomirsky
ParallelKittens: Systematic and Practical Simplification of Multi-GPU AI Kernels
Event Tensor: A Unified Abstraction for Compiling Dynamic Megakernel
AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimization
DynaFlow: Transparent and Flexible Intra-Device Parallelism via Programmable Operator Scheduling
ApproxMLIR : Accuracy-Aware Compiler for Compound ML System
HipKittens: Fast and Furious AMD Kernels
(ends 6:00 PM)
6 p.m.
Poster Session 3 [6:00-8:00]
Evergreen Ballroom
Posters -
ADR: AN AGENTIC DETECTION SYSTEMFORENTERPRISE AGENTIC AI SECURITY
PRISM: Parametrically Refactor Inference for Speculative Decoding Draft Models
PROMPTS: PeRformance Optimization via Multi-Agent Planning for LLM Training and Serving
CDLM: Consistency Diffusion Language Models for Faster Sampling
Agentic Operator Generation for ML ASICs
Speculative Decoding: Performance or Illusion?
Cost-aware Duration Prediction for Software Upgrades in Datacenters
SpecDiff-2: Scaling Diffusion Drafter Alignment For Faster Speculative Decoding
The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents
Accelerating Large-Scale Reasoning Model Inference with Sparse Self-Speculative Decoding
HELIOS : Adaptive Model And Early-Exit Selection for Efficient LLM Inference Serving
SHIP: SRAM-Based Huge Inference Pipelines for Fast LLM Serving
IntAttention: A Fully Integer Attention Pipeline for Efficient Edge Inference
BatchLLM: Optimizing Large Batched LLM Inference with Global Prefix Sharing and Throughput-oriented Token Batching
REPARO: LOSS-RESILIENT GENERATIVE CODEC FOR VIDEO CONFERENCING
Beyond the Buzz: A Pragmatic Take on Inference Disaggregation
Rethinking DVFS for Mobile LLMs: Unified Energy-Aware Scheduling with CORE
Optimizing Deployment Configurations for LLM Inference
EarthSight: A Distributed Framework for Low-Latency Satellite Intelligence
Meeting SLOs, Slashing Hours: Automated Enterprise LLM Optimization with OptiKIT
Scaling Up Large Language Models Serving Systems for Semantic Job Search
Sparing Strategies to Minimize Reliability Impact On Large Training Jobs
Massive-Scale Out-Of-Core UMAP on the GPU
Charon: A Unified and Fine-Grained Simulator for Large-Scale LLM Training and Inference
veScale-FSDP: Flexible and High-Performance FSDP at Scale
AXLearn: Modular, Hardware-Agnostic Large Model Training
Hawkeye: Reproducing GPU-Level Non-Determinism
GUARD: SCALABLE STRAGGLER DETECTION AND NODE HEALTH MANAGEMENT FOR LARGE-SCALE TRAINING
DriftBench: Measuring and Predicting Infrastructure Drift in LLM Serving Systems
MoEBlaze: Breaking the Memory Wall for Efficient MoE Training on Modern GPUs
PARROT: Persuasion and Agreement Robustness Rating of Output Truth — A Sycophancy Robustness Benchmark for LLMs
FreeScale: Distributed Training for Sequence Recommendation Models with Minimal Scaling Cost
ParallelKittens: Systematic and Practical Simplification of Multi-GPU AI Kernels
XProf: An Open, Scalable, and Extensible Profiling System for the Modern ML Stack
Event Tensor: A Unified Abstraction for Compiling Dynamic Megakernel
AIRS: Scaling Live Inference in Resource Constrained Environments
AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimization
SAKURAONE: An Open Ethernet–Based AI HPC System and Its Observed Workload Dynamics in a Single-Tenant LLM Development Environment
DynaFlow: Transparent and Flexible Intra-Device Parallelism via Programmable Operator Scheduling
MLCommons Chakra: Advancing Performance Benchmarking and Co-design using Standardized Execution Traces
ApproxMLIR : Accuracy-Aware Compiler for Compound ML System
ProfInfer: An eBPF-based Fine-Grained LLM Inference Profiler
HipKittens: Fast and Furious AMD Kernels
Machine Learning Fleet Efficiency: Improving TPU Systems at Scale with ML Productivity Goodput
Dataflow Is All You Need
Spira: Exploiting Voxel Data Structural Properties for Efficient Sparse Convolution in Point Cloud Networks
Efficient, VRAM-Constrained xLM Inference on Clients
Attribution-based Sparse Activation in Large Language Models
Wave: A Symbolic Python DSL And Compiler for High-Performance Machine Learning
Flashlight: PyTorch Compiler Extensions to Accelerate Attention Variants
CATWILD: Compiler Autotuning for TPU workloads in the Wild
FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling
(ends 8:00 PM)

FRI 22 MAY
8 a.m.
Registration
(ends 11:00 AM)
8:15 a.m.
Industry Track Oral Presentation: Compilers/HW [8:15-9:15]
Grand Ballroom 1
Moderator: Phitchaya Phothilimthana
Dataflow Is All You Need
Efficient, VRAM-Constrained xLM Inference on Clients
Wave: A Symbolic Python DSL And Compiler for High-Performance Machine Learning
CATWILD: Compiler Autotuning for TPU workloads in the Wild
(ends 9:15 AM)
Research Track Oral Presentation: Efficient Computation [8:15-9:15]
Grand Ballroom 2
Moderator: Aniruddh Ramrakhyani
Spira: Exploiting Voxel Data Structural Properties for Efficient Sparse Convolution in Point Cloud Networks
Attribution-based Sparse Activation in Large Language Models
Flashlight: PyTorch Compiler Extensions to Accelerate Attention Variants
FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling
(ends 9:15 AM)
9:15 a.m.
Break:
Coffee Break
(ends 9:45 AM)
9:45 a.m.
Keynote Talk:
The Path to Infernece Efficiency
Grand Ballroom 1
Christos Kozyrakis
(ends 10:45 AM)
10:45 a.m.
Break:
Coffee Break
(ends 11:00 AM)
11 a.m.
Competition:
Competition Track: Google / NVIDIA / Amazon
Grand Ballroom 1
(ends 12:50 PM)
1:30 p.m.