Heimdall automates translation of eBPF C programs to Rust with formal equivalence proofs for 94.1% of 102 tested programs using LLMs, static analysis, and Z3-based checking.
hub Mixed citations
Scaling symbolic evaluation for automated verification of systems code with serval
Mixed citation behavior. Most common role is background (43%).
21
Pith papers citing it
77
external citations · external index
Background
43% of classified citations
hub tools
citation-role summary
background 4
method 3
citation-polarity summary
representative citing papers
ForeMoE uses routing foresight from the rollout stage to enable micro-step load balancing in MoE RL post-training via a hierarchical planner and transfer engine, claiming up to 1.45x speedup on 64 GPUs.
RATrain introduces a resource-aware scheduler and MT-3000-specific backend for 1F1B LLM training that achieves 1.35x speedup and 97% scaling efficiency while preserving training correctness.
FlyCatcher infers 300 correct stateful runtime checkers from 400 tests across four systems, yielding 2.6x more correct checkers and 5.2x more error detections than prior work.
CTA-pipelining reduces 2-layer GEMM latency up to 31.8% vs micro-batching and 29.6% vs Tensor Parallelism on 8-GPU H200/B200 systems by CTA-level cross-GPU pipelining.
UBEP replaces BSP All-to-All for MoE on multi-tier superpods with dependency-driven kernel decomposition, topology-aware token scheduling, and Data-as-Flag atomics, cutting All-to-All latency up to 52.4% and TPOT up to 11.1%.
Entrain reduces microbatch workload variability by up to 10.6x and improves multimodal LLM training throughput by 1.4x via static model parallelism and deferred hierarchical microbatch assignment.
HyperParallel-MoE reduces Dispatch-to-Combine MoE-FFN latency by up to 1.58x on Ascend A3 clusters via tile-level heterogeneous scheduling that overlaps communication, matrix, and vector computation inside a single kernel launch.
Agentic model checking pairs LLM agents for spec inference and counterexample classification with compositional bounded model checking to verify LLM-generated kernel and compiler code.
A new tabular abstraction for pipeline schedules shows communication can reverse rankings from bubble analysis alone, with GPipe and 1F1B runtime-equivalent but 1F1B lower in activation memory.
RRFP introduces a readiness-driven runtime for pipeline parallelism that uses schedules as hints and ready-set arbitration to improve utilization under runtime variability, reporting up to 2.77x speedup on multimodal workloads.
ShardTensor is a domain-parallelism system for SciML that enables flexible scaling of extreme-resolution spatial datasets by removing the constraint of batch size one per device.
DBLP dynamically bounds gradient loss by training phase, cutting end-to-end training time ~24% on average and up to 5.88× single-round latency under microbursts versus a SOTA baseline.
Symphony detects step misalignments in ring collectives via lightweight in-network tracking and mitigates them by throttling outpacing flows with congestion signals, yielding up to 54% better communication times in Astra-Sim simulations and a Tofino2 prototype.
FM-Agent automates top-down natural-language Hoare reasoning with LLMs and reports 522 new bugs in large, previously tested systems up to 143k LoC.
R-DSGD and R-DSGD-M under (δ,c)-robust aggregation have tight Byzantine error floors under (B,ζ)-bounded dissimilarity; local momentum eliminates the stochastic-noise term but not the heterogeneity term.
SparseBalance dynamically adjusts sparsity and batches workloads to load-balance sparse attention training, delivering up to 1.33x speedup and 0.46% better long-context performance on LongBench.
Sampling parallelism distributes Bayesian sample evaluations across GPUs for near-perfect scaling, lower memory use, and faster convergence via per-GPU data augmentations, outperforming pure data parallelism in diversity.
FlexPipe introduces runtime pipeline refactoring for LLMs to achieve higher resource efficiency and lower latency in serverless GPU clusters with fragmentation.
NEURON-Fabric provides a profile-guided runtime for controlled low-bit gradient communication that preserves accuracy near full-precision levels while reducing modeled communication traffic across vision, transformer, and language model workloads.
ScaleAcross Explorer jointly optimizes three design dimensions for scale-across training and reports up to 64.62% speedups over production baselines and 37.59% over prior art in testbed and simulation experiments.
citing papers explorer
-
Heimdall: Formally Verified Automated Migration of Legacy eBPF Programs to Rust
Heimdall automates translation of eBPF C programs to Rust with formal equivalence proofs for 94.1% of 102 tested programs using LLMs, static analysis, and Z3-based checking.
-
Harnessing Routing Foresight for Micro-step-level MoE load balancing in RL Post-training
ForeMoE uses routing foresight from the rollout stage to enable micro-step load balancing in MoE RL post-training via a hierarchical planner and transfer engine, claiming up to 1.45x speedup on 64 GPUs.
-
RATrain: A Resource-Aware Training Runtime for Large Language Models on Bandwidth-Constrained Heterogeneous Supercomputing Platforms
RATrain introduces a resource-aware scheduler and MT-3000-specific backend for 1F1B LLM training that achieves 1.35x speedup and 97% scaling efficiency while preserving training correctness.
-
FlyCatcher: Neural Inference of Runtime Checkers from Tests
FlyCatcher infers 300 correct stateful runtime checkers from 400 tests across four systems, yielding 2.6x more correct checkers and 5.2x more error detections than prior work.
-
CTA-Pipelining: A Latency-Oriented Spatial Scaling Method for Multi-GPU Systems
CTA-pipelining reduces 2-layer GEMM latency up to 31.8% vs micro-batching and 29.6% vs Tensor Parallelism on 8-GPU H200/B200 systems by CTA-level cross-GPU pipelining.
-
UBEP: Re-architecting Expert Parallelism Communication Library for Production Superpods
UBEP replaces BSP All-to-All for MoE on multi-tier superpods with dependency-driven kernel decomposition, topology-aware token scheduling, and Data-as-Flag atomics, cutting All-to-All latency up to 52.4% and TPOT up to 11.1%.
-
Addressing Variable Heterogeneity in Distributed Multimodal Training with Entrain
Entrain reduces microbatch workload variability by up to 10.6x and improves multimodal LLM training throughput by 1.4x via static model parallelism and deferred hierarchical microbatch assignment.
-
HyperParallel-MoE: Multi-Core Interleaved Scheduling for Fast MoE Training on Ascend NPUs
HyperParallel-MoE reduces Dispatch-to-Combine MoE-FFN latency by up to 1.58x on Ascend A3 clusters via tile-level heterogeneous scheduling that overlaps communication, matrix, and vector computation inside a single kernel launch.
-
Agentic Model Checking
Agentic model checking pairs LLM agents for spec inference and counterexample classification with compositional bounded model checking to verify LLM-generated kernel and compiler code.
-
A Tabular Schedule Abstraction for Communication-Aware Evaluation of Pipeline-Parallel LLM Training
A new tabular abstraction for pipeline schedules shows communication can reverse rankings from bubble analysis alone, with GPipe and 1F1B runtime-equivalent but 1F1B lower in activation memory.
-
A Readiness-Driven Runtime for Pipeline-Parallel Training under Runtime Variability
RRFP introduces a readiness-driven runtime for pipeline parallelism that uses schedules as hints and ready-set arbitration to improve utilization under runtime variability, reporting up to 2.77x speedup on multimodal workloads.
-
ShardTensor: Domain Parallelism for Scientific Machine Learning
ShardTensor is a domain-parallelism system for SciML that enables flexible scaling of extreme-resolution spatial datasets by removing the constraint of batch size one per device.
-
DBLP: Phase-Aware Bounded-Loss Transport for Burst-Resilient Distributed ML Training
DBLP dynamically bounds gradient loss by training phase, cutting end-to-end training time ~24% on average and up to 5.88× single-round latency under microbursts versus a SOTA baseline.
-
Symphony: Taming Step Misalignments in the Network for Ring-based Collective Operations
Symphony detects step misalignments in ring collectives via lightweight in-network tracking and mitigates them by throttling outpacing flows with congestion signals, yielding up to 54% better communication times in Astra-Sim simulations and a Tofino2 prototype.
-
FM-Agent: Scaling Formal Methods to Large Systems via LLM-Based Hoare-Style Reasoning
FM-Agent automates top-down natural-language Hoare reasoning with LLMs and reports 522 new bugs in large, previously tested systems up to 143k LoC.
-
Tessera: Unlocking Heterogeneous GPUs through Kernel-Granularity Disaggregation
R-DSGD and R-DSGD-M under (δ,c)-robust aggregation have tight Byzantine error floors under (B,ζ)-bounded dissimilarity; local momentum eliminates the stochastic-noise term but not the heterogeneity term.
-
SparseBalance: Load-Balanced Long Context Training with Dynamic Sparse Attention
SparseBalance dynamically adjusts sparsity and batches workloads to load-balance sparse attention training, delivering up to 1.33x speedup and 0.46% better long-context performance on LongBench.
-
Sampling Parallelism for Fast and Efficient Bayesian Learning
Sampling parallelism distributes Bayesian sample evaluations across GPUs for near-perfect scaling, lower memory use, and faster convergence via per-GPU data augmentations, outperforming pure data parallelism in diversity.
-
FlexPipe: Adapting Dynamic LLM Serving Through Inflight Pipeline Refactoring in Fragmented Serverless Clusters
FlexPipe introduces runtime pipeline refactoring for LLMs to achieve higher resource efficiency and lower latency in serverless GPU clusters with fragmentation.
-
NEURON-Fabric: Architecture-Runtime Co-Design for Controlled Low-Bit Gradient Communication
NEURON-Fabric provides a profile-guided runtime for controlled low-bit gradient communication that preserves accuracy near full-precision levels while reducing modeled communication traffic across vision, transformer, and language model workloads.
-
ScaleAcross Explorer: Exploring Communication Optimization for Scale-Across AI Model Training
ScaleAcross Explorer jointly optimizes three design dimensions for scale-across training and reports up to 64.62% speedups over production baselines and 37.59% over prior art in testbed and simulation experiments.