Related papers: Louvre: Lightweight Ordering Using Versioning for …
Reinforcement Learning with Verifiable Rewards (RLVR) has advanced LLM reasoning, but remains constrained by inefficient exploration under limited rollout budgets, leading to low sampling success and unstable training in complex tasks. We…
Large Language Models (LLMs) remain vulnerable to prompt injection attacks, representing the most significant security threat in production deployments. We present Prompt Fencing, a novel architectural approach that applies cryptographic…
Many concurrent data-structure implementations use the well-known compare-and-swap (CAS) operation, supported in hardware by most modern multiprocessor architectures for inter-thread synchronization. A key weakness of the CAS operation is…
Multiprocess systems, including grid systems, multiprocessors and multicore computers, incorporate a variety of specialized hardware and software mechanisms, which speed computation, but result in complex memory behavior. As a consequence,…
Processing-in-memory (PIM) architectures allow software to explicitly initiate computation in the memory. This effectively makes PIM operations a new class of memory operations, alongside standard memory operations (e.g., load, store). For…
Application partitioning and code offloading are being researched extensively during the past few years. Several frameworks for code offloading have been proposed. However, fewer works attempted to address issues occurred with its…
Large language models now serve millions of users daily, with providers incurring costs exceeding $700,000 per day. Each request requires token-by-token inference, making GPU scheduling central to latency, capacity, and cost. The difficulty…
Memory consistency model (MCM) issues in out-of-order-issue microprocessor-based shared-memory systems are notoriously non-intuitive and a source of hardware design bugs. Prior hardware verification work is limited to in-order-issue…
We study efficient reasoning under tight compute. We ask how to make structured, correct decisions without increasing test time cost. We add two training only components to small and medium Transformers that also transfer to broader…
Naive backpropagation through time has a memory footprint that grows linearly in the sequence length, due to the need to store each state of the forward propagation. This is a problem for large networks. Strategies have been developed to…
Serverless computing has grown rapidly for serving Large Language Model (LLM) inference due to its pay-as-you-go pricing, fine-grained GPU usage, and rapid scaling. However, our analysis reveals that current serverless can effectively serve…
Large language model (LLM) agents on multi-step tasks suffer reasoning degradation, looping, drift, stuck states, at rates up to 30% on hard tasks. Current solutions include hard step limits (abrupt) or LLM-as-judge monitoring (10-15%…
Caches are used to reduce the speed differential between the CPU and memory to improve the performance of modern processors. However, attackers can use contention-based cache timing attacks to steal sensitive information from victim…
Systems programming often requires the manipulation of resources like file handles, network connections, or dynamically allocated memory. Programmers need to follow certain protocols to handle these resources correctly. Violating these…
Autoregressive models, despite their commendable performance in a myriad of generative tasks, face challenges stemming from their inherently sequential structure. Inference on these models, by design, harnesses a temporal dependency, where…
Modern deployment often requires trading accuracy for efficiency under tight CPU and memory constraints, yet common compression proxies such as parameter count or FLOPs do not reliably predict wall-clock inference time. In particular,…
We study the inherent space requirements of shared storage algorithms in asynchronous fault-prone systems. Previous works use codes to achieve a better storage cost than the well-known replication approach. However, a closer look reveals…
Due to the diversity and implicit redundancy in terms of processing units and compute kernels, off-the-shelf heterogeneous systems offer the opportunity to detect and tolerate faults during task execution in hardware as well as in software.…
In this paper, we propose a new approach to building synchronization primitives, dubbed "lwlocks" (short for light-weight locks). The primitives are optimized for small memory footprint while maintaining efficient performance in low…
The economics of prompt compression depend not only on reducing input tokens but on how compression changes output length, which is typically priced several times higher. We evaluate this in a pre-registered six-arm randomized controlled…