On-device research index

arXiv research

A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.

168,742 papers · 148 categories

Trend · papers per month

count % of corpus

12 results for “INT8”

Learning Accurate Integer Transformer Machine-Translation Modelscs.LG

Improved machine translation with INT8 hardware using a novel training method.

problem Training accurate machine translation models with limited hardware precision.

method Convert all Transformer matrix multiplications to 8-bit integer (INT8) without sacrificing accuracy.

result INT8 Transformer models achieve BLEU scores 99.3% to 100% relative to FP32 models.

machine translation Transformer models INT8 hardware matrix multiplications accuracy trade-offs

2020-01-03abs ↗pdf ↗

FrostNet: Towards Quantization-Aware Network Architecture Searchcs.LG

FrostNet improves INT8 quantization efficiency in mobile networks.

problem The importance of network architecture for optimal INT8 quantization.

method Quantization-aware training (QAT) with StatAssist and GradBoost, hardware-aware NAS.

result FrostNets achieve higher recognition accuracy with comparable latency when quantized.

INT8 quantization network architecture search quantization-aware training Frost bottleneck hardware-aware NAS

2020-06-17abs ↗pdf ↗

Efficient Residue Number System Based Winograd Convolutioncs.LG

Efficient Winograd convolution for INT8 networks using RNS.

problem Difficulty in applying Winograd algorithm to low-precision quantized networks.

method Extends Winograd algorithm to Residue Number System (RNS) for efficient INT8 convolution.

result Arithmetic complexity reduction up to 7.03x with performance improvement up to 2.30x-4.69x.

Winograd convolution Residue Number System INT8 Convolutional Neural Networks Low-Precision Arithmetic

2020-07-23abs ↗pdf ↗

Mapping the Schedule x Bit-Width Boundary in Sub-100M Quantisation-Aware Trainingcs.LG

Study finds optimal learning rate schedules for sub-100M quantization-aware training across bit-widths.

problem Optimal learning rate schedules for quantization-aware training depend on bit-width.

method Factorial grid testing over bit-width, warmdown fraction, LR magnitude, model size, and seed.

result INT6 QAT requires a different schedule than higher-precision training, falsifying the primary hypothesis.

quantization-aware training learning rate schedules sub-100M models bit-width warmdown fraction

2026-05-25abs ↗pdf ↗

Multi-Scale Dequant: Eliminating Dequantization Bottleneck via Activation Decomposition for Efficient LLM Inferencestat.ML

MSD removes dequantization bottleneck in LLM inference by approximating high-precision activations.

problem Dequantization bottleneck in LLM inference on modern AI accelerators.

method MSD decomposes high-precision activations into multiple low-precision components for direct multiplication with quantized weights.

result MSD avoids INT8-to-BF16 weight conversion, reducing dequantization cycles and HBM traffic.

dequantization bottleneck large language models quantization activation decomposition efficient inference

2026-05-13abs ↗pdf ↗

Degree-Quant: Quantization-Aware Training for Graph Neural Networkscs.LG

Degree-Quant improves GNN efficiency by quantizing them without losing accuracy.

problem Efficiency of graph neural networks at inference time.

method Architecturally-agnostic method, Degree-Quant, for quantizing GNNs.

result Degree-Quant trained models perform as well as full-precision models and achieve up to 26% gains.

Graph Neural Networks Quantization Efficiency Inference Integer Arithmetic

2020-08-11abs ↗pdf ↗

Quantization Aware Training, ERNIE and Kurtosis Regularizer: a short empirical studystat.ML

This study improves Ernie's accuracy for INT8 inference by modifying its training process.

problem Improving the accuracy of pre-trained models like Ernie for low precision inference.

method Integrates a regularizer into the training process to make it more robust to quantization.

result Increased INT8 accuracy for Ernie models.

Quantization Aware Training ERNIE Low Precision Inference Regularization Pre-trained Models

2021-06-24abs ↗pdf ↗

Once-for-All Channel Mixers (HYPERTINYPW): Generative Compression for TinyMLcs.LG

Generative compression technique reduces neural network size and improves performance on microcontrollers.

problem Memory constraints on microcontrollers limit the size of neural networks, especially for 1x1 pointwise (PW) mixers.

method HYPER-TINYPW uses a shared micro-MLP to generate PW kernels from tiny per-layer codes, reducing memory usage.

result HYPER-TINYPW achieves comparable performance to larger models while being significantly smaller (225 kB vs 1.4 MB).

compression microcontrollers neural networks 1x1 pointwise generative models

2026-03-26abs ↗pdf ↗

A Statistical Framework for Low-bitwidth Training of Deep Neural Networkscs.LG

Paper develops a statistical framework for quantized training of deep neural networks.

problem Lack of theoretical understanding of gradient quantization in FQT.

method Presented a statistical framework for analyzing FQT algorithms, viewing quantized gradient as a stochastic estimator of QAT gradient.

result Developed two novel gradient quantizers with smaller variance than existing per-tensor quantizer.

deep neural networks quantized training statistical framework gradient quantization ResNet-50

2020-10-27abs ↗pdf ↗

Training with Quantization Noise for Extreme Model Compressioncs.LG

We extend quantization-aware training to extreme model compression.

problem Maximizing model accuracy with minimal model size.

method Quantize a random subset of weights during training, allowing unbiased gradients through other weights.

result Established new state-of-the-art compromises between accuracy and model size.

model compression quantization straight-through estimator extreme compression accuracy vs. size

2020-04-15abs ↗pdf ↗

APQ: Joint Search for Network Architecture, Pruning and Quantization Policycs.LG

APQ jointly optimizes neural architecture, pruning, and quantization for efficient inference.

problem Efficient deep learning inference on resource-constrained hardware.

method Joint optimization of neural architecture, pruning, and quantization policy using a quantization-aware accuracy predictor.

result Joint optimization leads to 2.3% higher ImageNet accuracy with reduced latency and energy consumption.

deep learning inference optimization resource-constrained hardware joint optimization quantization

2020-06-15abs ↗pdf ↗

RFX: High-Performance Random Forests with GPU Acceleration and QLORA Compressioncs.LG

RFX accelerates and compresses Random Forests for large datasets.

problem Memory bottleneck in proximity matrices limits Random Forest analysis.

method QLORA compression, CPU TriBlock storage, GPU batch sizing, 3D MDS visualization.

result Proximity-based Random Forest analysis on larger datasets is feasible.

Random Forests GPU acceleration compression memory bottleneck quantization

2025-11-23abs ↗pdf ↗