arXiv subjects

George Karfakis

Publications and source records attributed to George Karfakis.

4 recordsLinked to original sources

arXiv2026

Enabling Retraining-Free Accurate Inference in Analog Compute in Memory Accelerators

Analog compute-in-memory (CIM) accelerators compute matrix-vector products directly inside memory arrays, eliminating weight movement and promising order-of-magnitude gains in inference efficiency. However, converter and device errors make pretrained models unusable on these arrays, and hardware-aware retraining is costly. We enable retraining-free, accurate inference of pretrained models on simple analog arrays with low-cost, shared calibration. Our key observation is that analog error depends on how each layer is represented on the array, and that this representation can be changed offline without changing the computation. We center the inputs and apply Hadamard rotations designed for analog arrays, which keep shared reference errors spread and confine input mixing to each array. Across CNNs, vision transformers and LLMs up to 14B parameters on PCM and ReRAM, the method brings vision models and larger LLMs within a few points of floating-point accuracy and matches or approaches published hardware-aware training without training any weights. On ReRAM with 6-bit ADCs, it restores ViT-B/16 from 37.4% to 82.9% and Phi-4 from 27.0% to 72.2%, both within 2.2 points of floating-point accuracy.

cs.AR↗

arXiv2026

RAPID-LLM: Resilience-Aware Performance analysis of Infrastructure for Distributed LLM Training and Inference

RAPID-LLM is a unified performance modeling framework for distributed large language model (LLM) training and inference on GPU clusters, without relying on deployment-specific traces or expensive cycle-level simulation for exploration. From a workload and hardware specification, it builds hardware-aware operator-level execution models that capture tiling, memory-hierarchy effects, communication, and memory feasibility under hybrid parallelism. Its backend simulates explicit multidimensional interconnects with congestion-aware routing and support for degraded and failed links, enabling scalable what-if analysis across topology, mapping, and hardware design choices. Across 124 evaluation cases spanning inference and dense, fully sharded, and mixture-of-experts training on A100 and H100 GPUs, RAPID-LLM achieves an overall mean absolute percentage error (MAPE) of 10.0\%. Its network predictions stay within 8\% of ns-3 on representative communication patterns. Case studies demonstrate how RAPID-LLM enables fast, systematic sweeps over hybrid-parallel configurations, quantifies sensitivity to link faults under realistic routing and congestion, and evaluates hypothetical GPU design variants including 3D-stacked HBM-on-GPU scenarios.

cs.PF↗

arXiv2026

MXFormer: A Microscaling Floating-Point Charge-Trap Transistor Compute-in-Memory Transformer Accelerator

The proliferation of Transformer models is often constrained by the significant computational and memory bandwidth demands of deployment. To address this, we present MXFormer, a novel, hybrid, weight-stationary Compute-in-Memory (CIM) accelerator that provides high throughput and efficiency for fixed-model inference on large short-sequence Transformers. Our architecture's foundation is the use of ultra-dense Charge-Trap Transistors (CTTs) in Microscaling MXFP4 CIM arrays, uniquely enabling the on-chip storage of up to hundreds of millions of parameters in Fully Weight Stationary (FWS) fashion. We introduce a statically partitioned design with 12 Transformer blocks connected by a deeply pipelined dataflow. Static-weight layers (MLPs and linear projections) execute on highly parallel analog CTT arrays using an MXFP4-native flow with per-block exponent alignment and a 10-bit SAR ADC. Dynamic computations are handled in fully accurate digital blocks that utilize MXFP-enabled systolic arrays for scaled dot-product attention and vector units for LayerNorm and FlashAttention-style Softmax. By eliminating all weight movement, the deeply pipelined MXFormer architecture yields very high single-stream throughput and efficiency, processing 58275 FPS on ViT-L/32 (dual-chip) or 41269 FPS on ViT-B/16 (single chip). MXFormer outperforms comparable state-of-the-art non-FWS digital, hybrid and photonic Transformer accelerators ~3.3x-60.5x in compute density and ~1.7x-2.5x in energy efficiency. Against FWS accelerators, MXFormer improves compute density by ~20.9x and resident weight storage density by ~2x, while preserving near-digital accuracy (drop of <1%) without any model retraining.

cs.AR↗

arXiv2023

A Search for Technosignatures Around 11,680 Stars with the Green Bank Telescope at 1.15-1.73 GHz

We conducted a search for narrowband radio signals over four observing sessions in 2020-2023 with the L-band receiver (1.15-1.73 GHz) of the 100 m diameter Green Bank Telescope. We pointed the telescope in the directions of 62 TESS Objects of Interest, capturing radio emissions from a total of ~11,680 stars and planetary systems in the ~9 arcminute beam of the telescope. All detections were either automatically rejected or visually inspected and confirmed to be of anthropogenic nature. In this work, we also quantified the end-to-end efficiency of radio SETI pipelines with a signal injection and recovery analysis. The UCLA SETI pipeline recovers 94.0% of the injected signals over the usable frequency range of the receiver and 98.7% of the injections when regions of dense RFI are excluded. In another pipeline that uses incoherent sums of 51 consecutive spectra, the recovery rate is ~15 times smaller at ~6%. The pipeline efficiency affects calculations of transmitter prevalence and SETI search volume. Accordingly, we developed an improved Drake Figure of Merit and a formalism to place upper limits on transmitter prevalence that take the pipeline efficiency and transmitter duty cycle into account. Based on our observations, we can state at the 95% confidence level that fewer than 6.6% of stars within 100 pc host a transmitter that is detectable in our search (EIRP > 1e13 W). For stars within 20,000 ly, the fraction of stars with detectable transmitters (EIRP > 5e16 W) is at most 3e-4. Finally, we showed that the UCLA SETI pipeline natively detects the signals detected with AI techniques by Ma et al. (2023).

astro-ph.IM↗