牛哥精选 · 三个月

1

🤖 AI·大模型 arXiv 机器学习 2026-05-25

Strong Teacher Not Needed? On Distillation in LLM Pretraining

颠覆认知？弱教师模型也能有效蒸馏LLM，预训练阶段教师强度并非关键。

arXiv:2605.23857v1 Announce Type: new Abstract: Knowledge distillation generally assumes a strong-to-weak relationship where stronger teachers yield b…

大语言模型知识蒸馏预训练模型压缩弱到弱蒸馏

2

📝 深度技术 arXiv AI 2026-05-23

Compiling Agentic Workflows into LLM Weights: Near-Frontier Quality at Two Orders of Magnitude Less Cost

将Agent工作流编译进LLM权重，以极低成本实现接近前沿的质量，提出了一种颠覆性的模型优化路径。

arXiv:2605.22502v1 Announce Type: new Abstract: Agent orchestration frameworks have proliferated, collectively exceeding 290,000 GitHub stars across L…

agentic wo llm权重编译成本优化模型压缩 agent编排

3

📝 深度技术 arXiv NLP 2026-05-22

X-Token: Projection-Guided Cross-Tokenizer Knowledge Distillation

无需辅助组件的投影引导跨分词器知识蒸馏，有效解决词汇不兼容问题。

arXiv:2605.21699v1 Announce Type: cross Abstract: Cross-tokenizer knowledge distillation allows a student model to learn from teachers with incompatib…

知识蒸馏跨分词器投影引导模型压缩学生模型

4

📝 深度技术 arXiv 机器学习 2026-05-20

LEAP: Learnable End-to-End Adaptive Pruning of Large Language Models

提出LEAP可学习端到端自适应剪枝方法，在保持大语言模型性能的同时实现高效压缩

arXiv:2605.17289v1 Announce Type: new Abstract: Unstructured sparsity is now natively accelerated by recent GPU kernels and dataflow hardware, shiftin…

leap 大语言模型自适应剪枝端到端可学习

5

🤖 AI·大模型 arXiv 机器学习 2026-05-20

Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training

混合全微调与低秩适应的新方法，专为后训练场景优化，效率与性能兼得

arXiv:2605.18822v1 Announce Type: new Abstract: Post-training has become essential for adapting large language models (LLMs) to complex downstream beh…

hybrid-lor 低秩适应全微调后训练模型压缩

6

📝 深度技术 arXiv NLP 2026-05-20

K-Quantization and its Impact on Output Performance

探索K-Quantization对模型输出性能的影响，量化新技术深度解析

arXiv:2605.19645v1 Announce Type: new Abstract: Recent advancements in large language models (LLMs) have shown their remarkable capacities in many NLP…

量化模型压缩输出性能 k-quantiza 深度学习

7

📝 深度技术 arXiv 机器学习 2026-05-20

Quantized Machine Learning Models for Medical Imaging in Low-Resource Healthcare Settings

量化技术让机器学习模型在低资源医疗影像场景下也能高效运行，大幅降低算力门槛，加速基层医疗智能化。

arXiv:2605.19207v1 Announce Type: cross Abstract: Deep learning models have shown strong performance in medical image analysis, but deploying them in …

量化模型医学影像低资源医疗模型压缩边缘部署

8

📝 深度技术 arXiv 机器学习 2026-05-20

Theory-optimal Quantization Based on Flatness

基于平坦度的理论最优量化方法，为深度学习模型压缩提供新思路

arXiv:2605.18800v1 Announce Type: new Abstract: Post-training quantization has emerged as a widely adopted technique for compressing and accelerating …

模型量化平坦度理论最优压缩深度学习 arxiv论文

9

📝 深度技术 arXiv NLP 2026-05-20

You Had One Job: Per-Task Quantization Using LLMs' Hidden Representations

利用大模型隐藏表示实现每任务量化，在保持性能的同时大幅提升效率，值得关注的技术突破。

arXiv:2511.06516v3 Announce Type: replace Abstract: Many LLM applications require only narrow capabilities, yet standard post-training quantization (P…

llm 量化隐藏表示模型压缩每任务量化

10

📝 深度技术 arXiv 机器学习 2026-05-20

RAP: Runtime Adaptive Pruning for LLM Inference

提出运行时自适应剪枝方法，让LLM推理内存动态调整，效率大增

arXiv:2505.17138v5 Announce Type: replace Abstract: Large language models (LLMs) excel at language understanding and generation, but their enormous co…

llm推理自适应剪枝运行时优化内存约束模型压缩

11

📝 深度技术 arXiv 机器学习 2026-05-20

Two-Valued Symmetric Circulant Matrices: Applications in Deep Learning

二值对称循环矩阵新结构，为深度学习在资源受限平台的高效部署提供突破性方案

arXiv:2605.16443v1 Announce Type: new Abstract: Despite the success of deep neural networks in vision, medical diagnosis, and IoT scenarios, their dep…

二值对称循环矩阵深度学习模型压缩资源受限部署矩阵理论

12

📝 深度技术 arXiv AI 2026-05-20

TAPIOCA: Why Task- Aware Pruning Improves OOD model Capability

探究任务感知剪枝如何提升模型在分布外数据上的表现，揭示内在机制

arXiv:2605.14738v1 Announce Type: cross Abstract: Recent work has promoted task-aware layer pruning as a way to improve model performance on particula…

任务感知剪枝 ood泛化多项式回归大语言模型模型压缩

13

📝 深度技术 arXiv 机器学习 2026-05-20

Prune, Update and Trim: Robust Structured Pruning for Large Language Models

提出新型结构化剪枝方法，实现大模型高效压缩同时保持鲁棒性，适合模型优化研究者

arXiv:2605.18331v1 Announce Type: new Abstract: Large Language Models (LLMs) have experienced significant growth and development in recent years. Howe…

大型语言模型结构化剪枝模型压缩鲁棒性剪枝方法

14

📝 深度技术 arXiv 机器学习 2026-05-19

GSQ: Highly-Accurate Low-Precision Scalar Quantization for LLMs via Gumbel-Softmax Sampling

提出GSQ方法，利用Gumbel-Softmax采样实现LLM的高精度低精度标量量化，突破现有量化瓶颈

arXiv:2604.18556v2 Announce Type: replace-cross Abstract: Quantization has become a standard tool for efficient LLM deployment, especially for local i…

llm量化 gumbel-sof 低精度量化标量量化模型压缩

15

📝 深度技术 arXiv 机器学习 2026-05-19

Algorithmic Simplification of Neural Networks with Mosaic-of-Motifs

新算法Mosaic-of-Motifs大幅简化神经网络，参数压缩性能损失极小，揭秘深度学习模型为何易于压缩。

arXiv:2602.14896v2 Announce Type: replace Abstract: Large-scale deep learning models are well-suited for compression. Across a variety of tasks, metho…

神经网络简化 mosaic-of- 模型压缩深度学习算法

16

📝 深度技术 arXiv 机器学习 2026-05-19

Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models

1-bit量化大模型新思路，输出对齐策略再审视，助力低资源设备高效推理

arXiv:2512.21651v3 Announce Type: replace Abstract: Large Language Models (LLMs) deliver strong performance across a wide range of NLP tasks, but thei…

llm 1-bit量化后训练量化输出对齐模型压缩

17

📝 深度技术 arXiv 机器学习 2026-05-19

IO-SVD: Input-Output Whitened SVD for Adaptive-Rank LLM Compression

提出输入输出白化SVD方法，实现自适应秩的大语言模型压缩，提升推理效率。

arXiv:2605.15626v1 Announce Type: new Abstract: Large language models deliver strong performance across language and reasoning tasks, but their storag…

svd 模型压缩低秩分解大语言模型预训练压缩

18

📝 深度技术 arXiv 机器学习 2026-05-19

Perforated Neural Networks for Keyword Spotting

一种新型穿孔神经网络，在关键词识别上同时提升准确率和压缩模型，突破边缘部署瓶颈。

arXiv:2605.15647v1 Announce Type: new Abstract: Edge machine learning presents a unique set of constraints not encountered in cloud-scale model deploy…

穿孔神经网络关键词识别边缘机器学习模型压缩准确性提升

🐂 牛哥精选

Strong Teacher Not Needed? On Distillation in LLM Pretraining

Compiling Agentic Workflows into LLM Weights: Near-Frontier Quality at Two Orders of Magnitude Less Cost

X-Token: Projection-Guided Cross-Tokenizer Knowledge Distillation

LEAP: Learnable End-to-End Adaptive Pruning of Large Language Models

Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training

K-Quantization and its Impact on Output Performance

Quantized Machine Learning Models for Medical Imaging in Low-Resource Healthcare Settings

Theory-optimal Quantization Based on Flatness

You Had One Job: Per-Task Quantization Using LLMs' Hidden Representations

RAP: Runtime Adaptive Pruning for LLM Inference

Two-Valued Symmetric Circulant Matrices: Applications in Deep Learning

TAPIOCA: Why Task- Aware Pruning Improves OOD model Capability

Prune, Update and Trim: Robust Structured Pruning for Large Language Models

GSQ: Highly-Accurate Low-Precision Scalar Quantization for LLMs via Gumbel-Softmax Sampling

Algorithmic Simplification of Neural Networks with Mosaic-of-Motifs

Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models

IO-SVD: Input-Output Whitened SVD for Adaptive-Rank LLM Compression

Perforated Neural Networks for Keyword Spotting

📅 日期