SOTAVerified

Quantization

Quantization is a promising technique to reduce the computation cost of neural network training, which can replace high-cost floating-point numbers (e.g., float32) with low-cost fixed-point numbers (e.g., int8/int16).

Source: Adaptive Precision Training: Quantify Back Propagation in Neural Networks with Fixed-point Numbers

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 101–150 of 4925 papers

TitleStatusHype
Fast Matrix Multiplications for Lookup Table-Quantized LLMsCode3
FlatQuant: Flatness Matters for LLM QuantizationCode3
Scaling Transformers for Low-Bitrate High-Quality Speech CodingCode3
APOLLO: SGD-like Memory, AdamW-level PerformanceCode3
RoSA: Accurate Parameter-Efficient Fine-Tuning via Robust AdaptationCode3
EfficientQAT: Efficient Quantization-Aware Training for Large Language ModelsCode3
A Survey on Inference Optimization Techniques for Mixture of Experts ModelsCode3
DPLM-2: A Multimodal Diffusion Protein Language ModelCode3
Pushing the Limits of Large Language Model Quantization via the Linearity TheoremCode3
Language-Codec: Bridging Discrete Codec Representations and Speech Language ModelsCode3
Ditto: Quantization-aware Secure Inference of Transformers upon MPCCode3
Data Generation for Hardware-Friendly Post-Training QuantizationCode3
PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language ModelsCode3
CV-VAE: A Compatible Video VAE for Latent Generative Video ModelsCode3
MuQ: Self-Supervised Music Representation Learning with Mel Residual Vector QuantizationCode3
MotionGPT: Human Motion as a Foreign LanguageCode3
NaturalSpeech 3: Zero-Shot Speech Synthesis with Factorized Codec and Diffusion ModelsCode3
8-bit Optimizers via Block-wise QuantizationCode3
BiLLM: Pushing the Limit of Post-Training Quantization for LLMsCode3
Addressing Representation Collapse in Vector Quantized Models with One Linear LayerCode3
FP6-LLM: Efficiently Serving Large Language Models Through FP6-Centric Algorithm-System Co-DesignCode3
MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of ExpertsCode3
OneBit: Towards Extremely Low-bit Large Language ModelsCode3
MAUVE Scores for Generative Models: Theory and PracticeCode2
Massive Values in Self-Attention Modules are the Key to Contextual Knowledge UnderstandingCode2
MBQ: Modality-Balanced Quantization for Large Vision-Language ModelsCode2
CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise ClassificationCode2
MC-MoE: Mixture Compressor for Mixture-of-Experts LLMs Gains MoreCode2
Accurate LoRA-Finetuning Quantization of LLMs via Information RetentionCode2
MAexp: A Generic Platform for RL-based Multi-Agent ExplorationCode2
4-bit Conformer with Native Quantization Aware Training for Speech RecognitionCode2
Low-Rank Quantization-Aware Training for LLMsCode2
MERT: Acoustic Music Understanding Model with Large-Scale Self-supervised TrainingCode2
LoftQ: LoRA-Fine-Tuning-Aware Quantization for Large Language ModelsCode2
LoQT: Low-Rank Adapters for Quantized PretrainingCode2
LoRANN: Low-Rank Matrix Factorization for Approximate Nearest Neighbor SearchCode2
LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPSCode2
LLM-FP4: 4-Bit Floating-Point Quantized TransformersCode2
Adapting Large Language Models by Integrating Collaborative Semantics for RecommendationCode2
LiDAR-PTQ: Post-Training Quantization for Point Cloud 3D Object DetectionCode2
Lossless Compression of Vector IDs for Approximate Nearest Neighbor SearchCode2
MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group QuantizationCode2
BMInf: An Efficient Toolkit for Big Model Inference and TuningCode2
Bolt: Accelerated Data Mining with Fast Vector CompressionCode2
KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark of Long Context Capable ApproachesCode2
I-ViT: Integer-only Quantization for Efficient Vision Transformer InferenceCode2
BitNet: Scaling 1-bit Transformers for Large Language ModelsCode2
BitVLA: 1-bit Vision-Language-Action Models for Robotics ManipulationCode2
INT-FlashAttention: Enabling Flash Attention for INT8 QuantizationCode2
Jetfire: Efficient and Accurate Transformer Pretraining with INT8 Data Flow and Per-Block QuantizationCode2
Show:102550
← PrevPage 3 of 99Next →
All datasetsImageNetCIFAR-10Wiki-40BAgeDB-30CFP-FPCOCO (Common Objects in Context)IJB-BIJB-CKnowledge-based:LFW

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1FQ-ViT (ViT-L)Top-1 Accuracy (%)85.03—Unverified
2FQ-ViT (ViT-B)Top-1 Accuracy (%)83.31—Unverified
3FQ-ViT (Swin-B)Top-1 Accuracy (%)82.97—Unverified
4FQ-ViT (Swin-S)Top-1 Accuracy (%)82.71—Unverified
5FQ-ViT (DeiT-B)Top-1 Accuracy (%)81.2—Unverified
6FQ-ViT (Swin-T)Top-1 Accuracy (%)80.51—Unverified
7FQ-ViT (DeiT-S)Top-1 Accuracy (%)79.17—Unverified
8Xception W8A8Top-1 Accuracy (%)78.97—Unverified
9ADLIK-MO-ResNet50-W4A4Top-1 Accuracy (%)77.88—Unverified
10ADLIK-MO-ResNet50-W3A4Top-1 Accuracy (%)77.34—Unverified
#ModelMetricClaimedVerifiedStatus
13DCNN_VIVA_3MAP160,327.04—Unverified
2DTQMAP0.79—Unverified
#ModelMetricClaimedVerifiedStatus
1OutEffHop-Bert_basePerplexity6.3—Unverified
2OutEffHop-Bert_basePerplexity6.21—Unverified
#ModelMetricClaimedVerifiedStatus
1Accuracy98.13—Unverified
#ModelMetricClaimedVerifiedStatus
1Accuracy92.92—Unverified
#ModelMetricClaimedVerifiedStatus
1SSD ResNet50 V1 FPN 640x640MAP34.3—Unverified
#ModelMetricClaimedVerifiedStatus
1TAR @ FAR=1e-495.13—Unverified
#ModelMetricClaimedVerifiedStatus
1TAR @ FAR=1e-496.38—Unverified
#ModelMetricClaimedVerifiedStatus
13DCNN_VIVA_5All84,809,664—Unverified
#ModelMetricClaimedVerifiedStatus
1Accuracy99.8—Unverified