1
测量精度突破1微米，效率提升3倍，优可测高精度闪测仪发布
国产闪测仪精度突破0.7微米，效率提升3倍，打破国外高端仪器卡脖子困境，赋能精密制造。
2
PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs
针对长思维链大语言模型的KV缓存量化新突破，渐进混合精度方案，已被ICLR 2026接收，代码已开源。
arXiv:2505.18610v2 Announce Type: replace Abstract: Recently, significant progress has been made in developing reasoning-capable Large Language Models…
3
黑爵推出 AK039 磁轴单手键盘：38 键布局 / 8K 回报率 / RT0.01 精度，到手 199 元
IT之家 7 月 11 日消息，黑爵推出了一款 AK039 磁轴单手键盘，定价为 199 元。该产品为电竞级磁轴小键盘，采用 38 键布局加金属旋钮设计，相比全键盘更节省空间，可最大化鼠标移动范围，适合作为游戏第二把键盘。 京东 黑爵（AJAZZ）AK039 磁轴键盘 199 元 直达链接 该键盘搭…
4
支撑 Level 4 自动驾驶：曝特斯拉 Cybercab 出租车支持双频 GPS
特斯拉为无人驾驶出租车配双频GPS，精度提升支撑Level 4自动驾驶。
IT之家 7 月 10 日消息，科技媒体 notateslaapp 今天（7 月 10 日）发布博文，报道称特斯拉为提升无人驾驶出租车 Cybercab 的定位精度与冗余能力，将为其配备双频 GPS 方案。 IT之家援引博文介绍，为支撑 SAE Level 4 自动驾驶能力，特斯拉为这款无人驾驶出租…
5
GIFT: Geometry-Informed Low-precision Gradient Communication for LLM Pretraining
提出GIFT方法，利用梯度几何信息实现低精度通信，在不牺牲模型精度的前提下显著降低LLM预训练通信开销。
arXiv:2607.07494v1 Announce Type: cross Abstract: Gradient communication is a primary scaling bottleneck in large language model (LLM) pretraining. Co…
6
HiFA4: Training-Free 4-bit FlashAttention on Ascend HIF4 NPUs for LLM Inference
华为昇腾NPU上实现免训练4-bit FlashAttention，大幅降低LLM推理内存与计算开销。
arXiv:2607.04302v1 Announce Type: cross Abstract: We present HiFA4, a post-training operator-level design that executes both QK^T and PV in FlashAtten…
7
填补国内空白：我国建成首套高精度圆度基准装置，圆度测量不确定度从 20nm 降至 6nm
IT之家 7 月 7 日消息，据央视新闻昨日报道，我国首套高精度圆度基准装置正式建成。 据介绍，该装置由中国计量科学研究院研制完成，其建成标志着我国在高精度圆度测量领域实现了跨越式发展，整体技术能力跻身国际先进行列。 IT之家注：圆度是几何量形位公差体系中的核心基础参数之一，其测量精度直接关系到精密…
8
LACUNA: A Testbed for Evaluating Localization Precision for LLM Unlearning
新测试平台LACUNA精准评估大模型遗忘方法，定位精度是关键突破点。
arXiv:2607.02513v1 Announce Type: cross Abstract: LLMs memorize sensitive training data, including personally identifiable information (PII), creating…
9
ATK 梵塔推出 FUZZY 63V2/60V2 磁轴键盘：0.001mm RT 精度、8kHz 轮询率，到手价 1299 元
IT之家 7 月 1 日消息，ATK 梵塔今日推出 FUZZY 63V2 标准版和 FUZZY 60V2 标准版磁轴键盘，将于 7 月 8 日 20 点开售， 首发到手价 1299 元 。 其中，63V2 标准版为左移 64% 配列，可选 6 系铝合金定位板和编织碳纤维定位板；60V2 标准版为 6…
10
CAT-Q: Cost-efficient and Accurate Ternary Quantization for LLMs
被ICML 2026接收为oral，提出一种兼顾成本与精度的三元量化新方法，为大模型高效部署提供突破。
arXiv:2606.26650v1 Announce Type: cross Abstract: In this paper, we present CAT-Q, Cost-efficient and Accurate Ternary Quantization, for compressing a…
11
Enabling Cloud-Level Accuracy in Edge AI through IoT Data Preprocessing
这篇arXiv论文提出通过IoT数据预处理让边缘AI达到云级精度，值得技术研究者关注。
arXiv:2606.22496v1 Announce Type: cross Abstract: Large language models (LLMs) offer a natural-language interface for interpreting Internet of Things …
12
Rethinking Shrinkage Bias in LLM FP4 Pretraining: Geometric Origin, Systemic Impact, and UFP4 Recipe
直击FP4低精度训练痛点，首次从几何角度解释收缩偏差的根源，并提出UFP4方案系统性解决该问题。
arXiv:2606.20381v1 Announce Type: new Abstract: FP4 training promises substantial reductions in memory and computation cost for LLM pretraining, yet c…
13
TileFuse: A Fused Mixed-Precision Kernel Library for Efficient Quantized LLM Inference on AMD NPUs
面向AMD NPU的融合混合精度内核库，显著提升量化大语言模型推理效率。
arXiv:2606.11357v1 Announce Type: cross Abstract: With the growing demand for on-device LLM inference, edge SoCs increasingly integrate NPUs to improv…
14
Trainable Smooth-Rotation Transforms with Learned Channel Scales for LLM Quantization
提出可训练的平滑旋转变换与学习通道缩放，为LLM量化精度提升提供新思路
arXiv:2606.09927v1 Announce Type: cross Abstract: Post-training quantization (PTQ) is one of the most practical ways to reduce the serving cost of Lar…
15
ScaleSweep: Accurate NVFP4 Post-Training Quantization of LLMs via Block Scale Initialization
提出基于块尺度初始化的NVFP4后训练量化方法，有效提升大语言模型低比特精度。
arXiv:2606.07618v1 Announce Type: new Abstract: NVFP4 is a recently introduced hardware-supported FP4 format that improves the fidelity of 4-bit quant…
16
The Routing Plateau: Understanding and Breaking the Accuracy Limits of LLM Routers
LLM路由技术遭遇精度瓶颈？21种方法深度研究揭示突破之道。
arXiv:2606.07587v1 Announce Type: new Abstract: LLM routing has become a popular approach to improve the cost-quality trade-off of LLM services by dyn…
17
Joint Structural Pruning and Mixed-Precision Quantization for LLM Compression
联合结构化剪枝与混合精度量化，为LLM压缩提供高效新方案，兼顾精度与速度。
arXiv:2606.07819v1 Announce Type: cross Abstract: Recently, the efficiency of Large Language Models (LLMs) deployment has become a critical concern in…
18
GGUF vs. GPTQ vs. AWQ: The Plain-English Guide to LLM Quantization
一文讲清GGUF、GPTQ、AWQ三种主流LLM量化格式的区别，帮你选对压缩方式省显存。
Article URL: https://vettedconsumer.com/gguf-vs-gptq-vs-awq-the-plain-english-guide-to-llm-quantization-and-which-one-to-pick/ Comments URL: https://n…
19
Show HN: LLM memory without context bleed; 100% precision vs. <10% vector search
终结LLM上下文泄露：100%精度记忆，告别向量搜索的10%低效。
Article URL: https://tenureai.dev/ Comments URL: https://news.ycombinator.com/item?id=48409678 Points: 4 # Comments: 1
20
Channel-Wise Mixed-Precision Quantization for Large Language Models
大语言模型量化新突破：通道级混合精度策略如何在保持性能同时压缩模型？
arXiv:2410.13056v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have demonstrated remarkable success across a wide range of lan…