Research Papers
Research Papers news today
21 stories from 2 sources in the last 48 hours, every one in English. Each links to its publisher.
-
Melting and resistivity of iron at Earth core conditions using pulsed Joule heating
Science Magazine Tech
Science, Ahead of Print.
-
ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
arXiv cs.DC (Distributed)
arXiv:2607.00466v3 Announce Type: replace Abstract: Equal load does not mean equal decode latency in prefill-decode (PD) disaggregated mixture-of-experts (MoE) serving. In bandwidth-bound decode, more distinct experts mean more weight…
-
Online Graph Exploration under T-Interval Connectivity
arXiv cs.DC (Distributed)
arXiv:2604.04619v2 Announce Type: replace Abstract: We study deterministic exploration by a single agent in $T$-interval-connected graphs, a standard model of dynamic networks in which, for every time window of length $T$, the intersection…
-
TPIC: Energy-Conserving Space Plasma Simulations on GPU Supercomputers
arXiv cs.DC (Distributed)
arXiv:2611.09624v1 Announce Type: cross Abstract: Artificial heating and numerical damping affect particle energization in kinetic space plasma simulations, even when the numerical scheme remains stable. Implicit and semi implicit energy…
-
PaB-tree: Parallel B+-tree with Batch Updates
arXiv cs.DC (Distributed)
arXiv:2611.09172v1 Announce Type: cross Abstract: The B+-tree is a ubiquitous search index due to its provable I/O (cache) efficiency. Yet unlike batch-parallel binary search trees, which are both fast and widely available, no…
-
Fast and Actionable Classification of HPC I/O Time Series
arXiv cs.DC (Distributed)
arXiv:2611.09646v1 Announce Type: new Abstract: Input/output (I/O) is a major bottleneck for highperformance computing (HPC) performance. One of the challenges is the contention between jobs when sharing the parallel file system. However…
-
Nebula: Distributed Position-Independent Caching for LLM Serving
arXiv cs.DC (Distributed)
arXiv:2611.09474v1 Announce Type: new Abstract: Modern LLM workflows, such as retrieval-augmented generation (RAG) and LLM agents, build long prompts from multiple independent text segments, making prefill a major source of computation…
-
Green500: Adventure in Qualifying a Cluster
arXiv cs.DC (Distributed)
arXiv:2611.09251v1 Announce Type: new Abstract: Energy effciency has become a key metric for evaluating High Performance Computing (HPC) systems, as reflected by the Green500 ranking. However, energy effciency alone does not capture the…
-
DARIC: A Position-Independent KV Cache Reuse System with Distribution-Aware Recomputation
arXiv cs.DC (Distributed)
arXiv:2611.09109v1 Announce Type: new Abstract: The context lengths of large language models (LLMs) are rapidly increasing in real-world applications such as agentic systems and retrieval-augmented generation (RAG). To accelerate LLM…
-
Inside the Trap: A Renewal Law for Prefix-Cache Collapse in LLM Agent Serving
arXiv cs.DC (Distributed)
arXiv:2611.08902v1 Announce Type: new Abstract: Prefix caching lets an LLM agent's follow-up turns reuse cached KV, but a burst can tip an engine into a trap in which every turn misses and the backlog does not drain, at a load it served…
-
Introducing Cluster Efficiency to LLM Serving: Pooling Economical Accelerators over Fast Interconnects
arXiv cs.DC (Distributed)
arXiv:2611.02975v2 Announce Type: replace Abstract: Long-context large language model (LLM) serving places substantial demands on compute, memory capacity, and memory bandwidth. Economical accelerators can offer these resources at lower…
-
ResServe: Accurate and Efficient FP4 LLM Inference with Residue Compensation
arXiv cs.DC (Distributed)
arXiv:2611.08713v1 Announce Type: new Abstract: Block-scaled 4-bit floating-point (FP4) quantization reduces large language model inference costs but can degrade task quality. Higher-precision computation can mitigate this loss at the…
-
Greedy-Like Defective Coloring: Distributed Algorithms and Applications
arXiv cs.DC (Distributed)
arXiv:2608.02386v3 Announce Type: replace Abstract: A $d$-defective $c$-coloring of a graph $G=(V,E)$ is a coloring of the nodes $V$ with $c$ colors such that every node has at most $d$ neighbors of the same color. Distributed algorithms…
-
Load Balancing with Item Fragmentation and More Bins than Items: Linear-Time and Exact
arXiv cs.DC (Distributed)
arXiv:2611.09267v1 Announce Type: cross Abstract: We study a fundamental load-balancing problem in which $N$ items of strictly positive integer sizes with total size $W$ must be assigned to $B$ bins, with items allowed to be divided into…
-
Concurrent Use of FP32 and FP64 Computing Units on GPUs and Mixed-Precision Placement in WENO Reconstruction
arXiv cs.DC (Distributed)
arXiv:2611.09651v1 Announce Type: new Abstract: Scientific computing often requires FP64 arithmetic, although FP64 performance is limited on NVIDIA GPUs. Double-float (DF) emulation, which represents a value using two FP32 values with…
-
Towards Sustainable Federated Learning via Carbon-Aware Client Selection
arXiv cs.DC (Distributed)
arXiv:2611.09552v1 Announce Type: new Abstract: The Information/Communication Technology (ICT) industry is severely challenging the global environment due to the sharply rising carbon emissions generated by machine learning (ML) workloads…
-
UniExit: A Unified Latency-Aware Early-Exit Runtime for Heterogeneous Edge AI
arXiv cs.DC (Distributed)
arXiv:2611.09159v1 Announce Type: new Abstract: Edge applications increasingly combine CNN-based perception with edge-deployable Transformer-based language models, but deploying such heterogeneous workloads on resource-constrained devices…
-
ISER-MoE: Dynamic In-Switch Replication of Hot Experts for Load Balancing in Expert-Parallel
arXiv cs.DC (Distributed)
arXiv:2611.09008v1 Announce Type: new Abstract: Expert-parallel mixture-of-experts (MoE) layers run at the pace of the most loaded GPU, and a few hot experts carry most of the excess load. Replicating them at every MoE layer balances the…
-
The GPU Needs an Operating System
arXiv cs.DC (Distributed)
arXiv:2611.08720v1 Announce Type: new Abstract: AI chips sit mostly idle while serving models, because replies must be fast and each step waits on memory. On two large open models, the math units were busy only about 6-11% of the time. We…
-
AutoParallel: Composable, PyTorch-Native Automatic Parallelism for Production Training
arXiv cs.DC (Distributed)
arXiv:2611.08660v1 Announce Type: new Abstract: Training large models requires multiple interacting forms of parallelism. As models and clusters scale, finding and maintaining efficient parallelization plans becomes essential for training…
-
ReKV: Overlapping Quantized KV Transfer and Suffix Recomputation
arXiv cs.DC (Distributed)
arXiv:2611.08770v1 Announce Type: new Abstract: Prefill--decode disaggregation moves a prompt's key--value (KV) cache between workers before the first token can be generated. Reducing this handoff to either cache compression or…