论文

arxiv:2609.25270
复制 Markdown
RULER：用于 SVG 生成的实例感知量规奖励
发布于 9月21日
·
提交者
Hangyu Ran
于 9月23日

今日第2篇论文

·
inclusionAI
作者：
Hangyu Ran
,
Yuhao Zheng
,
Yingying Zhang
,
Kevin Qinghong Lin
,
Han Peng
摘要

从自然语言指令生成可缩放矢量图形（SVG）代码是一个开放-ended任务，缺乏绝对的视觉真实值，导致评估和策略优化都缺乏可靠的信号。在自然图像上校准的标量指标（CLIP、美学评分）难以迁移到风格化的矢量内容，且将其作为强化学习奖励会引发奖励黑客攻击。我们通过基于量规的评分来解决这两个局限性。我们首先通过实证确立，使用多维量规提示视觉语言评判者，其结果与人类判断的相关性远高于标量指标，无论是在样本间还是指令内均如此。基于这一发现，我们引入了 RULER（用于强化学习的实例感知量规奖励），它将每条指令转换为涵盖语义、视觉和风格维度的六项实例感知量规；评判者 VLM 逐项对渲染出的 rollout 进行评分，加权后的满意度通过组相对策略优化形成细粒度奖励。由于量规仅从文本推导得出，RULER 既不需要配对的 SVG 真实值，也不需要人类偏好标签。在 MMSVG-Illustration 和 MMSVG-Icon 数据集上，RULER 将量规评分从 0.432/0.395 提升至 0.693/0.683，超越了专门的 SVG 专家模型，并达到了与规模大得多的 DeepSeek-V3 相当的水平；消融实验表明，量规设计是开放-ended SVG 生成中强化学习的活跃杠杆。项目页面位于 https://hangyuran.github.io/RULER/。

查看 arXiv 页面
查看 PDF
项目页面
添加到合集
社区
hangyuran
论文作者
论文提交者
大约 20 小时前

项目页面：https://hangyuran.github.io/RULER/
代码：将于 https://github.com/ant-research/RULER 发布

查看翻译
🔥
3
+
回复
researchstudio-bot
大约 16 小时前

这是来自 ResearchStudio 团队的自动消息。

我们为此论文创建了一个交互式的 ResearchStudio Reel。它包括视觉海报、视频和博客，所有文件均可下载且格式可编辑。

打开 ResearchStudio Reel →

从 Hugging Face 下载所有文件

如果您觉得该 Reel 有帮助，请给此评论点赞！

想探索或为更多论文创建 Reel？请访问 ResearchStudio 演示页面。

查看翻译
回复
gogo8232
大约 2 小时前
•
编辑于大约 2 小时前

[空] 关于这项研究的工作。然而，在阅读您的工作之前，我想提醒您该名称已被占用了一段时间。请查看以下内容。

https://github.com/NVIDIA/RULER

查看翻译
回复
编辑
预览
通过拖拽、粘贴或点击此处上传图片、音频和视频。
评论

· 注册或登录以发表评论

赞同
59
+47

在您的智能体中获取此论文：

hf papers read 2609.25270
没有最新的 CLI？
引用此论文的模型
0

无链接此论文的模型

在模型的 README.md 中引用 arxiv.org/abs/2609.25270 以从此页面链接它。
引用此论文的数据集
0

无链接此论文的数据集

在数据集的 README.md 中引用 arxiv.org/abs/2609.25270 以从此页面链接它。
引用此论文的 Spaces
0

无链接此论文的 Space

在 Space 的 README.md 中引用 arxiv.org/abs/2609.25270 以从此页面链接它。
包含此论文的合集
0

无包含此论文的合集

将此论文添加到合集中以从此页面链接它。

Papers
arxiv:2609.25270
Copy markdown
RULER: Instance-aware Rubric Rewards for SVG Generation
Published on Sep 21
·
Submitted by
Hangyu Ran
on Sep 23

2 Paper of the day

·
inclusionAI
Authors:
Hangyu Ran
,
Yuhao Zheng
,
Yingying Zhang
,
Kevin Qinghong Lin
,
Han Peng
Abstract

Generating Scalable Vector Graphics (SVG) code from natural-language instructions is an open-ended task without absolute visual ground truth, leaving both evaluation and policy optimization without a faithful signal. Scalar metrics (CLIP, Aesthetic) calibrated on natural images transfer poorly to stylized vector content, and reusing them as RL rewards triggers reward hacking. We address both limitations with rubric-based scoring. We first establish empirically that prompting a vision-language judge with a multi-axis rubric correlates with human judgments far better than scalar metrics, both across samples and within instructions. Building on this finding, we introduce RULER (Instance-aware Rubric Rewards for Reinforcement Learning), which converts each instruction into an instance-aware rubric of six items spanning semantic, visual, and stylistic axes; a judge VLM scores rendered rollouts item-by-item, and the weighted satisfactions form a fine-grained reward optimized via Group Relative Policy Optimization. Because the rubric is derived from text alone, RULER requires neither paired SVG ground truth nor human preference labels. On MMSVG-Illustration and MMSVG-Icon, RULER lifts the rubric score from 0.432/0.395 to 0.693/0.683, surpassing dedicated SVG specialists and matching the substantially larger DeepSeek-V3, with ablations identifying rubric design as the active lever for RL on open-ended SVG generation. The project page is available at https://hangyuran.github.io/RULER/.

View arXiv page
View PDF
Project page
Add to collection
Community
hangyuran
Paper author
Paper submitter
about 20 hours ago

Project Page: https://hangyuran.github.io/RULER/
Code: Will be released at https://github.com/ant-research/RULER

See translation
🔥
3
+
Reply
researchstudio-bot
about 16 hours ago

This is an automated message from the ResearchStudio team.

We created an interactive ResearchStudio Reel for this paper. It includes a visual poster, a video, and a blog, all available for download in editable formats.

Open the ResearchStudio Reel →

Download all files from Hugging Face

Please give this comment a thumbs up if you find the Reel helpful!

Want to explore or create Reels for more papers? Visit the ResearchStudio demo.

See translation
Reply
gogo8232
about 2 hours ago
•
edited about 2 hours ago

[empty] job on the research. However, before reading your work, I would like to warn you that the name was taken a while ago. check out the below.

https://github.com/NVIDIA/RULER

See translation
Reply
Edit
Preview
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Comment

· Sign up or log in to comment

Upvote
59
+47

Get this paper in your agent:

hf papers read 2609.25270
Don't have the latest CLI?
Models citing this paper
0

No model linking this paper

Cite arxiv.org/abs/2609.25270 in a model README.md to link it from this page.
Datasets citing this paper
0

No dataset linking this paper

Cite arxiv.org/abs/2609.25270 in a dataset README.md to link it from this page.
Spaces citing this paper
0

No Space linking this paper

Cite arxiv.org/abs/2609.25270 in a Space README.md to link it from this page.
Collections including this paper
0

No Collection including this paper

Add this paper to a collection to link it from this page.