A single fused int4 KV cache kernel on Apple Silicon outperforms fp16 in latency with 3x memory compression and near-zero quality loss on tested models.
Title resolution pending
17 Pith papers cite this work. Polarity classification is still indexing.
17
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
QuantVLA is the first post-training quantization framework for VLA models that quantizes the diffusion transformer action head and reports higher task success rates than full-precision baselines with roughly 70% memory savings on the quantized components.
Four Over Six adaptively scales blocks in NVFP4 quantization to smaller FP4 values, making representable value distributions more uniform and reducing quantization error especially for near-maximal values.
ARHQ isolates error-sensitive weight directions in LLMs via truncated SVD on the scaled matrix W G_x^{1/2} from activation residuals, improving SNR and preserving performance under aggressive low-bit quantization.
MemExplorer optimizes heterogeneous memory systems for agentic LLM inference on NPUs and reports up to 2.3x higher energy efficiency than baselines under fixed power budgets.
Redefining residual errors to include compensation-aware discrepancies and realigning calibration to full-precision outputs improves GPTQ and GPTAQ performance on LLMs.
Double achieves up to 5.3x inference speedup on 70B LLMs via synchronous double retrieval speculative parallelism that is lossless and outperforms trained baselines like EAGLE-3.
TAQ estimates per-layer importance from hidden representations and output sensitivity on task calibration data to allocate mixed precision in a training-free PTQ setting, outperforming task-agnostic baselines on accuracy-memory ratio across benchmarks.
TurboQuant achieves near-optimal vector quantization distortion for both MSE and inner products via random rotation and per-coordinate scalar quantization, with a formal proof that it matches lower bounds within a factor of approximately 2.7.
A ReRAM-on-logic stacked chip delivers 14.08-135.69 tokens/s LLM inference with block-clustered compression and adaptive parallel speculative decoding, yielding 4.46-7.17x speedup over standard methods.
HeadQ applies score-space logit corrections for keys and attention-weighted surrogates for values to KV-cache quantization, removing 84-94% of excess perplexity in 2-bit key experiments across six models.
Chunked streaming top-k enables CSA indexer execution at 1M sequence length with 6.21 GB peak memory and >=0.998 recall on synthetic V4-shaped inputs.
High-rate quantization theory yields accurate approximations for the distortion of absmax INT and FP schemes in generic weight-plus-activation matrix multiplication.
PLENA introduces a co-designed system with three optimization pathways for long-context agentic LLM inference, claiming up to 2.23x throughput over A100 and 4.04x energy efficiency.
MixLLM uses global output-feature importance to set mixed bit-widths for LLM quantization and adds two-step dequantization plus software pipelining for system efficiency.
DuQuant++ adapts outlier-aware fine-grained rotation to MXFP4 by matching block size to the 32-element microscaling group, enabling a single rotation that smooths distributions and achieves SOTA performance on LLaMA-3 with lower cost.
The paper surveys techniques to speed up and reduce the resource needs of LLM inference, organized by data-level, model-level, and system-level changes, with comparative experiments on representative methods.
citing papers explorer
-
When Quantization Is Free: An int4 KV Cache That Outruns fp16 on Apple Silicon
A single fused int4 KV cache kernel on Apple Silicon outperforms fp16 in latency with 3x memory compression and near-zero quality loss on tested models.
-
QuantVLA: Scale-Calibrated Post-Training Quantization for Vision-Language-Action Models
QuantVLA is the first post-training quantization framework for VLA models that quantizes the diffusion transformer action head and reports higher task success rates than full-precision baselines with roughly 70% memory savings on the quantized components.
-
Four Over Six: More Accurate NVFP4 Quantization with Adaptive Block Scaling
Four Over Six adaptively scales blocks in NVFP4 quantization to smaller FP4 values, making representable value distributions more uniform and reducing quantization error especially for near-maximal values.
-
Technical Report: Activation Residual Hessian Quantization (ARHQ) for Low-Bit LLM Quantization
ARHQ isolates error-sensitive weight directions in LLMs via truncated SVD on the scaled matrix W G_x^{1/2} from activation residuals, improving SNR and preserving performance under aggressive low-bit quantization.
-
MemExplorer: Navigating the Heterogeneous Memory Design Space for Agentic Inference NPUs
MemExplorer optimizes heterogeneous memory systems for agentic LLM inference on NPUs and reports up to 2.3x higher energy efficiency than baselines under fixed power budgets.
-
Rethinking Residual Errors in Compensation-based LLM Quantization
Redefining residual errors to include compensation-aware discrepancies and realigning calibration to full-precision outputs improves GPTQ and GPTAQ performance on LLMs.
-
Double: Breaking the Acceleration Limit via Double Retrieval Speculative Parallelism
Double achieves up to 5.3x inference speedup on 70B LLMs via synchronous double retrieval speculative parallelism that is lossless and outperforms trained baselines like EAGLE-3.
-
You Had One Job: Per-Task Quantization Using LLMs' Hidden Representations
TAQ estimates per-layer importance from hidden representations and output sensitivity on task calibration data to allocate mixed precision in a training-free PTQ setting, outperforming task-agnostic baselines on accuracy-memory ratio across benchmarks.
-
TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate
TurboQuant achieves near-optimal vector quantization distortion for both MSE and inner products via random rotation and per-coordinate scalar quantization, with a formal proof that it matches lower bounds within a factor of approximately 2.7.
-
31.1 A 14.08-to-135.69Token/s ReRAM-on-Logic Stacked Outlier-Free Large-Language-Model Accelerator with Block-Clustered Weight-Compression and Adaptive Parallel-Speculative-Decoding
A ReRAM-on-logic stacked chip delivers 14.08-135.69 tokens/s LLM inference with block-clustered compression and adaptive parallel speculative decoding, yielding 4.46-7.17x speedup over standard methods.
-
HeadQ: Model-Visible Distortion and Score-Space Correction for KV-Cache Quantization
HeadQ applies score-space logit corrections for keys and attention-weighted surrogates for values to KV-cache quantization, removing 84-94% of excess perplexity in 2-bit key experiments across six models.
-
StreamIndex: Memory-Bounded Compressed Sparse Attention via Streaming Top-k
Chunked streaming top-k enables CSA indexer execution at 1M sequence length with 6.21 GB peak memory and >=0.998 recall on synthetic V4-shaped inputs.
-
High-Rate Quantized Matrix Multiplication I
High-rate quantization theory yields accurate approximations for the distortion of absmax INT and FP schemes in generic weight-plus-activation matrix multiplication.
-
Combating the Memory Walls: Optimization Pathways for Long-Context Agentic LLM Inference
PLENA introduces a co-designed system with three optimization pathways for long-context agentic LLM inference, claiming up to 2.23x throughput over A100 and 4.04x energy efficiency.
-
MixLLM: LLM Quantization with Global Mixed-precision between Output-features and Highly-efficient System Design
MixLLM uses global output-feature importance to set mixed bit-widths for LLM quantization and adds two-step dequantization plus software pipelining for system efficiency.
-
DuQuant++: Fine-grained Rotation Enhances Microscaling FP4 Quantization
DuQuant++ adapts outlier-aware fine-grained rotation to MXFP4 by matching block size to the 32-element microscaling group, enabling a single rotation that smooths distributions and achieves SOTA performance on LLaMA-3 with lower cost.
-
A Survey on Efficient Inference for Large Language Models
The paper surveys techniques to speed up and reduce the resource needs of LLM inference, organized by data-level, model-level, and system-level changes, with comparative experiments on representative methods.