-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathnews.html
More file actions
177 lines (169 loc) · 11.4 KB
/
Copy pathnews.html
File metadata and controls
177 lines (169 loc) · 11.4 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
<!DOCTYPE html>
<html lang="en" data-lang="en">
<head>
<meta charset="UTF-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>News — FutureMLS Lab</title>
<meta name="description" content="News and updates from the Future Machine Learning & Systems (FutureMLS) Lab — releases, milestones, and publications." />
<link rel="icon" type="image/png" href="assets/img/logo.png" />
<!-- Choose the starting language before first paint. A saved choice always wins;
otherwise Chinese-locale or mainland-China visitors start in Chinese. -->
<script>
(function () {
var lang;
try { lang = localStorage.getItem("lang"); } catch (e) {}
if (lang !== "en" && lang !== "zh") {
var zh = false;
try {
zh = (navigator.languages || [navigator.language || ""]).some(function (l) {
return /^zh/i.test(l);
}) || /^Asia\/(Shanghai|Chongqing|Harbin|Urumqi)$/.test(
Intl.DateTimeFormat().resolvedOptions().timeZone || ""
);
} catch (e) {}
lang = zh ? "zh" : "en";
}
if (lang === "zh") {
document.documentElement.dataset.lang = "zh";
document.documentElement.lang = "zh-CN";
}
})();
</script>
<link rel="preload" href="assets/fonts/source-serif-4-latin.woff2" as="font" type="font/woff2" crossorigin />
<link rel="preload" href="assets/fonts/inter-latin.woff2" as="font" type="font/woff2" crossorigin />
<link rel="stylesheet" href="assets/css/style.css" />
</head>
<body>
<!-- ===== Navigation ===== -->
<header class="nav" id="nav">
<div class="nav__inner container">
<a class="nav__brand" href="index.html" aria-label="FutureMLS Lab home">
<img src="assets/img/logo.png" alt="" class="nav__logo" />
<span class="nav__brand-text">FutureMLS <span data-i18n="en">Lab</span><span data-i18n="zh">实验室</span></span>
</a>
<nav class="nav__links" id="navLinks" aria-label="Primary">
<a href="index.html"><span data-i18n="en">Home</span><span data-i18n="zh">首页</span></a>
<a href="news.html" aria-current="page"><span data-i18n="en">News</span><span data-i18n="zh">新闻</span></a>
<a href="https://github.com/FutureMLS-Lab" class="nav__ext" target="_blank" rel="noopener">GitHub</a>
</nav>
<button class="nav__toggle" id="navToggle" aria-label="Toggle menu" aria-expanded="false">
<span></span><span></span><span></span>
</button>
</div>
</header>
<main class="proj container">
<a class="proj__back" href="index.html#home"><span data-i18n="en">← Back to home</span><span data-i18n="zh">← 返回首页</span></a>
<span class="eyebrow" data-i18n="en">News</span>
<span class="eyebrow" data-i18n="zh">新闻</span>
<h1 class="proj__title" data-i18n="en">News & Updates</h1>
<h1 class="proj__title" data-i18n="zh">新闻与动态</h1>
<p class="proj__lead" data-i18n="en">
Releases, milestones, and publications from the Future Machine Learning & Systems Lab.
</p>
<p class="proj__lead" data-i18n="zh">
来自 Future Machine Learning & Systems 实验室的发布、里程碑与论文动态。
</p>
<ul class="news">
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">Jun 26, 2026</span>
<span class="news__cat" data-i18n="en">Project</span>
<span class="news__cat" data-i18n="zh">项目</span>
</div>
<p class="news__text" data-i18n="en">Happy to see community interest in bringing <strong><a href="projects/oscar.html">OSCAR</a></strong> toward vLLM serving workflows. Grateful to the contributors pushing broader support for deployable rotation-based KV-cache quantization in mainstream LLM serving stacks.</p>
<p class="news__text" data-i18n="zh">很高兴看到社区正在把 <strong><a href="projects/oscar.html">OSCAR</a></strong> 推向 vLLM 的服务链路。感谢各位贡献者,让基于旋转的可部署 KV-cache 量化在主流 LLM 服务栈中获得更广泛的支持。</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">Jun 19, 2026</span>
<span class="news__cat" data-i18n="en">Project</span>
<span class="news__cat" data-i18n="zh">项目</span>
</div>
<p class="news__text" data-i18n="en"><strong><a href="projects/oscar.html">OSCAR</a></strong> has been covered by multiple tech media outlets, including Towards AI, ModelScope, MarkTechPost, QbitAI (量子位), and Synced (机器之心), reaching 100,000+ reads across platforms. Grateful for the community interest in deployable 2-bit KV-cache quantization.</p>
<p class="news__text" data-i18n="zh"><strong><a href="projects/oscar.html">OSCAR</a></strong> 被多家科技媒体报道,包括 Towards AI、ModelScope、MarkTechPost、量子位与机器之心,全平台阅读量超过 100,000 次。感谢社区对可部署 2-bit KV-cache 量化的关注。</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">Jun 18, 2026</span>
<span class="news__cat" data-i18n="en">Project</span>
<span class="news__cat" data-i18n="zh">项目</span>
</div>
<p class="news__text" data-i18n="en">Released <strong><a href="projects/taylor-calibrate.html">Taylor-Calibrate</a></strong>: Principled Initialization for Hybrid Linear Attention Distillation. The codebase distills Qwen/Llama softmax-attention Transformers into hybrid linear-attention students built on GatedDeltaNet, using Taylor-series-informed initialization before staged distillation.</p>
<p class="news__text" data-i18n="zh">发布 <strong><a href="projects/taylor-calibrate.html">Taylor-Calibrate</a></strong>:面向混合线性注意力蒸馏的原理化初始化方法。该代码库把 Qwen/Llama 的 softmax attention Transformer 蒸馏为基于 GatedDeltaNet 的混合线性注意力学生模型,并在分阶段蒸馏前先完成基于泰勒展开的初始化。</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">Jun 15, 2026</span>
<span class="news__cat news__cat--paper" data-i18n="en">Paper</span>
<span class="news__cat news__cat--paper" data-i18n="zh">论文</span>
</div>
<p class="news__text" data-i18n="en">Our paper <strong><a href="projects/oscar.html">OSCAR</a>: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization</strong> has been accepted by <strong>ACL SELVA 2026</strong>!</p>
<p class="news__text" data-i18n="zh">我们的论文 <strong><a href="projects/oscar.html">OSCAR</a>: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization</strong> 已被 <strong>ACL SELVA 2026</strong> 接收!</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">Jun 10, 2026</span>
<span class="news__cat" data-i18n="en">Project</span>
<span class="news__cat" data-i18n="zh">项目</span>
</div>
<p class="news__text" data-i18n="en">Huge thanks to the open-source community — <strong><a href="projects/oscar.html">OSCAR</a></strong> has reached 500 stars on GitHub and now supports local llama.cpp usage, making 2-bit KV-cache quantization easier to try on local LLM deployments.</p>
<p class="news__text" data-i18n="zh">衷心感谢开源社区——<strong><a href="projects/oscar.html">OSCAR</a></strong> 的 GitHub star 数已达 500,并且现已支持在本地 llama.cpp 中使用,让 2-bit KV-cache 量化更容易在本地 LLM 部署上试用。</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">May 26, 2026</span>
<span class="news__cat" data-i18n="en">Project</span>
<span class="news__cat" data-i18n="zh">项目</span>
</div>
<p class="news__text" data-i18n="en"><strong><a href="projects/oscar.html">OSCAR</a></strong> crosses 300 ★ on GitHub in its first week — thanks to the open-source community.</p>
<p class="news__text" data-i18n="zh"><strong><a href="projects/oscar.html">OSCAR</a></strong> 上线首周 GitHub star 数突破 300 ★——感谢开源社区。</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">May 19, 2026</span>
<span class="news__cat" data-i18n="en">Project</span>
<span class="news__cat" data-i18n="zh">项目</span>
</div>
<p class="news__text" data-i18n="en"><strong><a href="projects/oscar.html">OSCAR</a></strong> released — 2-bit KV-cache serving at 2.28 effective bits/element with near-BF16 accuracy on Qwen3 and GLM-4.7.</p>
<p class="news__text" data-i18n="zh"><strong><a href="projects/oscar.html">OSCAR</a></strong> 正式发布——在 Qwen3 与 GLM-4.7 上以 2.28 等效 bit/元素实现 2-bit KV-cache 服务,精度接近 BF16。</p>
</li>
<li class="news__item reveal">
<div class="news__meta">
<span class="news__date">Apr 2026</span>
<span class="news__cat news__cat--paper" data-i18n="en">Paper</span>
<span class="news__cat news__cat--paper" data-i18n="zh">论文</span>
</div>
<p class="news__text" data-i18n="en"><strong><a href="projects/care.html">CARE</a></strong> presented at <strong>ICLR 2026</strong>: covariance-aware, rank-enhanced decomposition for Multi-Head Latent Attention.</p>
<p class="news__text" data-i18n="zh"><strong><a href="projects/care.html">CARE</a></strong> 在 <strong>ICLR 2026</strong> 发表:面向 Multi-Head Latent Attention 的协方差感知、秩增强分解方法。</p>
</li>
</ul>
</main>
<!-- ===== Footer ===== -->
<footer class="footer">
<div class="container footer__inner">
<div class="footer__brand">
<img src="assets/img/logo.png" alt="" class="footer__logo" />
<div>
<strong>FutureMLS <span data-i18n="en">Lab</span><span data-i18n="zh">实验室</span></strong>
<p>Future Machine Learning & Systems <span data-i18n="en">Lab</span><span data-i18n="zh">实验室</span></p>
</div>
</div>
<nav class="footer__links" aria-label="Footer">
<a href="index.html#about"><span data-i18n="en">About</span><span data-i18n="zh">关于</span></a>
<a href="news.html"><span data-i18n="en">News</span><span data-i18n="zh">新闻</span></a>
<a href="index.html#team"><span data-i18n="en">Team</span><span data-i18n="zh">团队</span></a>
<a href="index.html#projects"><span data-i18n="en">Projects</span><span data-i18n="zh">项目</span></a>
<a href="https://github.com/FutureMLS-Lab" target="_blank" rel="noopener">GitHub</a>
</nav>
<p class="footer__copy">© <span id="year">2026</span>
<span data-i18n="en">Future Machine Learning & Systems Lab. Founded by Zhongzhu Zhou.</span>
<span data-i18n="zh">Future Machine Learning & Systems 实验室。由周中柱创立。</span>
</p>
</div>
</footer>
<button class="langtoggle" id="langToggle" type="button" aria-label="Switch language / 切换语言">
<span data-i18n="en">中文</span><span data-i18n="zh">English</span>
</button>
<script src="assets/js/main.js"></script>
</body>
</html>