KV cache quantization silently erodes LLM safety alignment via vulnerable low-dimensional subspaces, diagnosed by Per-Channel Reduction into three failure modes and mitigated training-free with up to 97% recovery.
hub Canonical reference
Smoothquant: Accurate and efficient post-training quantization for large language models
Canonical reference. 100% of citing Pith papers cite this work as background.
33
Pith papers citing it
Background
100% of classified citations
hub tools
citation-role summary
background 7
citation-polarity summary
roles
background 7polarities
background 7representative citing papers
LlamaWeb is a WebGPU backend for llama.cpp that uses static memory planning, tunable kernels, and templated multi-precision support to cut memory use by 29-33% and raise decode throughput by 45-69% versus prior browser frameworks on tested hardware.
FP32-converged language models enter a post-convergence phase where INT4 quantization error explodes while FP32 perplexity remains stable, with onset tied to fine convergence rather than learning rate decay.
CORP performs one-shot structured pruning of Transformers by modeling removed components as affine functions of retained ones and solving closed-form ridge regressions on calibration data to fold compensation into weights, retaining 83.27% Top-1 accuracy on DeiT-Huge after 50% pruning.
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.
QLoRA finetunes 4-bit quantized LLMs via LoRA adapters to match full-precision performance while using far less memory, enabling 65B-scale training on single GPUs and producing Guanaco models near ChatGPT level.
SharQ combines input-adaptive N:M sparsity and FP4 quantization via sparse backbone plus dense residual, recovering 43-63% of the NVFP4-to-FP16 accuracy gap on Llama and Qwen models without calibration or retraining.
GRINQH introduces a graded input-based quantization hierarchy that dynamically assigns multi-precision weights using activation magnitudes as importance proxy, unifying quantization with sparsification to improve LLM decoding speed and quality trade-offs on Llama3 and Qwen3 models.
HyperQuant unifies Hadamard transform, optimal lattice quantization, and entropy coding to outperform prior schemes on LLM weight and KV cache quantization down to 1.7 bits per scalar while preserving quality on a 19B DiT model.
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
MCWC aligns permutation-symmetric blocks across layers to enable sequential prediction and residual entropy coding, improving rate-accuracy tradeoffs versus quantization and prior codecs on language and vision models.
Sign-flip perturbations produce π/(π-2) ≈ 2.75 times more transverse output energy than equal-norm sign-preserving perturbations in a ReLU + RMSNorm block because ReLU creates directional asymmetry that RMSNorm's transverse projection exposes.
LoKA enables practical FP8 use in numerically sensitive large recommendation models via online profiling of activations, reusable model modifications for stability, and dynamic kernel dispatching.
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.
MCAP uses load-time Monte Carlo profiling to estimate layer importance, enabling dynamic quantization (W4A8 vs W4A16) and memory tiering (GPU/RAM/SSD) that delivers 1.5-1.8x higher decode throughput than llama-cpp Q4_0 on NVIDIA T4 while fitting models into previously infeasible memory budgets.
Sol-RL decouples FP4-based candidate exploration from BF16 policy optimization in diffusion RL, delivering up to 4.64x faster convergence with maintained or superior alignment performance on models like FLUX.1 and SD3.5.
H2O evicts non-heavy-hitter tokens from the KV cache using a dynamic submodular policy, retaining recent and frequent-co-occurrence tokens to reduce memory while preserving accuracy.
AWQ quantizes LLM weights to low bits by scaling salient channels based on activation statistics, outperforming prior methods on language, coding, math, and multi-modal benchmarks.
FrugalGPT learns query-specific cascades across heterogeneous LLM APIs to match or exceed top-model accuracy at far lower cost.
SAGE-PTQ is a graph-guided ultra-low-bit PTQ framework that achieves 1.03 average weight bits and 0.004 scaling bits per matrix on LLMs while reporting lower perplexity and memory use than BiLLM and PB-LLM.
Cassandra is a self-speculative decoding system that builds a draft model via fine-grained data selection and optimized pruning/mantissa truncation, achieving up to 2.41x speedup over BF16 and 1.81x more tokens than Eagle-3 on Llama 3 8B without training.
TACO compresses tensor-parallel intermediate tensors with an adaptive FP8 scheme and fused kernels, yielding up to 1.87X throughput gains on GPT and Qwen models with near-lossless accuracy.
BloomBee is a distributed LLM inference system that achieves up to 1.76x higher throughput and 43.2% lower latency than prior decentralized systems by optimizing communication across multiple dimensions in low-bandwidth internet settings.
KL divergence provides a superior forward-only metric for identifying quantization-sensitive parts in SSM-Transformer hybrids, outperforming MSE and SQNR and supporting practical mixed-precision deployment on edge devices.
citing papers explorer
-
Alignment Collapse Under KV Cache Quantization: Diagnosis and Mitigation
KV cache quantization silently erodes LLM safety alignment via vulnerable low-dimensional subspaces, diagnosed by Per-Channel Reduction into three failure modes and mitigated training-free with up to 97% recovery.
-
Llamas on the Web: Memory-Efficient, Performance-Portable, and Multi-Precision LLM Inference with WebGPU
LlamaWeb is a WebGPU backend for llama.cpp that uses static memory planning, tunable kernels, and templated multi-precision support to cut memory use by 29-33% and raise decode throughput by 45-69% versus prior browser frameworks on tested hardware.
-
When Flat Minima Fail: Characterizing INT4 Quantization Collapse After FP32 Convergence
FP32-converged language models enter a post-convergence phase where INT4 quantization error explodes while FP32 perplexity remains stable, with onset tied to fine convergence rather than learning rate decay.
-
CORP: Closed-Form One-shot Representation-Preserving Structured Pruning for Transformers
CORP performs one-shot structured pruning of Transformers by modeling removed components as affine functions of retained ones and solving closed-form ridge regressions on calibration data to fold compensation into weights, retaining 83.27% Top-1 accuracy on DeiT-Huge after 50% pruning.
-
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.
-
QLoRA: Efficient Finetuning of Quantized LLMs
QLoRA finetunes 4-bit quantized LLMs via LoRA adapters to match full-precision performance while using far less memory, enabling 65B-scale training on single GPUs and producing Guanaco models near ChatGPT level.
-
SharQ: Bridging Activation Sparsity and FP4 Quantization for LLM Inference
SharQ combines input-adaptive N:M sparsity and FP4 quantization via sparse backbone plus dense residual, recovering 43-63% of the NVFP4-to-FP16 accuracy gap on Llama and Qwen models without calibration or retraining.
-
GRINQH: Graded Input-based Quantization Hierarchy for Efficient LLM Generation
GRINQH introduces a graded input-based quantization hierarchy that dynamically assigns multi-precision weights using activation magnitudes as importance proxy, unifying quantization with sparsification to improve LLM decoding speed and quality trade-offs on Llama3 and Qwen3 models.
-
HyperQuant: A Rate-Distortion-Optimal Quantization Pipeline for Large Language and Diffusion Models
HyperQuant unifies Hadamard transform, optimal lattice quantization, and entropy coding to outperform prior schemes on LLM weight and KV cache quantization down to 1.7 bits per scalar while preserving quality on a 19B DiT model.
-
A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
-
Motion-Compensated Weight Compression
MCWC aligns permutation-symmetric blocks across layers to enable sequential prediction and residual entropy coding, improving rate-accuracy tradeoffs versus quantization and prior codecs on language and vision models.
-
A Geometric Analysis of Sign-Magnitude Asymmetry in a ReLU + RMSNorm Block under Ternary Quantization
Sign-flip perturbations produce π/(π-2) ≈ 2.75 times more transverse output energy than equal-norm sign-preserving perturbations in a ReLU + RMSNorm block because ReLU creates directional asymmetry that RMSNorm's transverse projection exposes.
-
LoKA: Low-precision Kernel Applications for Recommendation Models At Scale
LoKA enables practical FP8 use in numerically sensitive large recommendation models via online profiling of activations, reusable model modifications for stability, and dynamic kernel dispatching.
-
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.
-
MCAP: Deployment-Time Layer Profiling for Memory-Constrained LLM Inference
MCAP uses load-time Monte Carlo profiling to estimate layer importance, enabling dynamic quantization (W4A8 vs W4A16) and memory tiering (GPU/RAM/SSD) that delivers 1.5-1.8x higher decode throughput than llama-cpp Q4_0 on NVIDIA T4 while fitting models into previously infeasible memory budgets.
-
FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling
Sol-RL decouples FP4-based candidate exploration from BF16 policy optimization in diffusion RL, delivering up to 4.64x faster convergence with maintained or superior alignment performance on models like FLUX.1 and SD3.5.
-
H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models
H2O evicts non-heavy-hitter tokens from the KV cache using a dynamic submodular policy, retaining recent and frequent-co-occurrence tokens to reduce memory while preserving accuracy.
-
AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
AWQ quantizes LLM weights to low bits by scaling salient channels based on activation statistics, outperforming prior methods on language, coding, math, and multi-modal benchmarks.
-
FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance
FrugalGPT learns query-specific cascades across heterogeneous LLM APIs to match or exceed top-model accuracy at far lower cost.
-
Minimizing the Hidden Cost of Scales: Graph-Guided Ultra-Low-Bit Quantization for Large Language Models
SAGE-PTQ is a graph-guided ultra-low-bit PTQ framework that achieves 1.03 average weight bits and 0.004 scaling bits per matrix on LLMs while reporting lower perplexity and memory use than BiLLM and PB-LLM.
-
Cassandra: Enabling Reasoning LLMs at Edge via Self-Speculative Decoding
Cassandra is a self-speculative decoding system that builds a draft model via fine-grained data selection and optimized pruning/mantissa truncation, achieving up to 2.41x speedup over BF16 and 1.81x more tokens than Eagle-3 on Llama 3 8B without training.
-
TACO: Efficient Communication Compression of Intermediate Tensors for Scalable Tensor-Parallel LLM Training
TACO compresses tensor-parallel intermediate tensors with an adaptive FP8 scheme and fused kernels, yielding up to 1.87X throughput gains on GPT and Qwen models with near-lossless accuracy.
-
Distributed Generative Inference of LLM at Internet Scales with Multi-Dimensional Communication Optimization
BloomBee is a distributed LLM inference system that achieves up to 1.76x higher throughput and 43.2% lower latency than prior decentralized systems by optimizing communication across multiple dimensions in low-bandwidth internet settings.
-
A KL Lens on Quantization: Fast, Forward-Only Sensitivity for Mixed-Precision SSM-Transformer Models
KL divergence provides a superior forward-only metric for identifying quantization-sensitive parts in SSM-Transformer hybrids, outperforming MSE and SQNR and supporting practical mixed-precision deployment on edge devices.
-
ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference
ShadowNPU presents shadowAttn, a co-designed sparse attention system that uses NPU pilot compute and techniques like graph bucketing and per-head sparsity to minimize CPU/GPU fallback during on-device LLM inference while maintaining accuracy.
-
$\text{Log}_\text{b}$Quant: Quantizing Language Models in Logarithmic Space
Log_b Quant is an adjustable-base logarithmic quantization technique that outperforms tensor-wise asymmetric linear quantization at 4-bit precision on language model benchmarks while providing memory savings.
-
Holding the FP8 Quality Ceiling at 8-Bit Weights and Activations: INT8 and GGUF Post-Training Quantization of Ideogram 4.0 for Consumer GPUs
INT8 W8A8 post-training quantization of Ideogram 4.0 preserves FP8 quality on a 200-prompt benchmark while outperforming NF4 on CLIP score and offering a favorable quality-memory trade-off via GGUF Q4_K.
-
Influence-Inspired Spectral Rotations for Extreme Low-Bit LLM Quantization
A WHT rotation plus per-coordinate activation-energy rescaling before auto-round quantization lowers WikiText-2 perplexity 15-58% versus vanilla auto-round at W2A16 on models from 135M to 1.5B parameters.
-
Token-Operations-Oriented Inference Optimization Techniques for Large Models
The paper introduces a four-layer technical architecture for token-operations-oriented inference optimization in large models and reviews key technologies and industry status at each layer.
-
Collaborative Few-Step Distillation and Low-Bit Quantization for Wan2.2 Dual-Expert Video Diffusion Models
A co-designed few-step distillation and low-bit quantization pipeline for Wan2.2-T2V-A14B keeps quantized few-step performance close to or above the full-precision baseline at 8 and 20 steps.
-
The Hitchhiker's Guide to Agentic AI: From Foundations to Systems
A comprehensive reference book organizing existing techniques for agentic AI systems across LLM substrate, reasoning, agent design patterns, inter-agent coordination, and production deployment.
- Attention Sink Forges Native MoE in Attention Layers: Sink-Aware Training to Address Head Collapse
- You Had One Job: Per-Task Quantization Using LLMs' Hidden Representations