---
格式版本: 2
标题: "MARCO: Click-Intent Decomposition for Calibrated Ads Conversion Prediction"
原文链接: "https://arxiv.org/abs/2608.10562"
发布日期: "2026-08-11"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Tue, 11 Aug 2026 06:49:59 UTC (631 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-12T11:17:29+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-12T11:13:54+08:00"
入库时间: "2026-08-12T03:17:29.987Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=deployment&searchtype=all"
匹配关键词:
  - "deployment"
相关厂家:
  []
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 0
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "内容为arXiv上关于广告点击转化预测的机器学习论文，完全未涉及超节点/AI Rack/机柜级AI基础设施、高速互连、供电、散热、量产落地等任何项目关注领域，与主题无关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-12T11:17:48+08:00"
AI主题相关性: 0
AI来源权威性: 0
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
采集批次: "2026年8月12日2点04分20秒"
采集批次ID: "20260812-020420-289"
去重键: "https://arxiv.org/abs/2608.10562"
---

## Computer Science > Machine Learning

## Title:MARCO: Click-Intent Decomposition for Calibrated Ads Conversion Prediction

Authors:[Shiwen Shen](https://arxiv.org/search/cs?searchtype=author&query=Shen,+S), [Xiru Huang](https://arxiv.org/search/cs?searchtype=author&query=Huang,+X), [Liang Luo](https://arxiv.org/search/cs?searchtype=author&query=Luo,+L), [Jianbo Sun](https://arxiv.org/search/cs?searchtype=author&query=Sun,+J), [He Lyu](https://arxiv.org/search/cs?searchtype=author&query=Lyu,+H), [Zihang Fu](https://arxiv.org/search/cs?searchtype=author&query=Fu,+Z), [Ivonne Xu](https://arxiv.org/search/cs?searchtype=author&query=Xu,+I), [Zhizhuo Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+Z), [Zhengyu Zhang](https://arxiv.org/search/cs?searchtype=author&query=Zhang,+Z), [Pei-Ju Sung](https://arxiv.org/search/cs?searchtype=author&query=Sung,+P), [Yunmiao Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+Y), [Zixuan Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+Z), [Zhengli Zhao](https://arxiv.org/search/cs?searchtype=author&query=Zhao,+Z), [Qiang Jin](https://arxiv.org/search/cs?searchtype=author&query=Jin,+Q), [Mike Jermann](https://arxiv.org/search/cs?searchtype=author&query=Jermann,+M), [Mingda Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+M), [Yang Xiao](https://arxiv.org/search/cs?searchtype=author&query=Xiao,+Y), [Bhavana Challa](https://arxiv.org/search/cs?searchtype=author&query=Challa,+B), [Brooke Bian](https://arxiv.org/search/cs?searchtype=author&query=Bian,+B), [Yang Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+Y), [Ashish Chamoli](https://arxiv.org/search/cs?searchtype=author&query=Chamoli,+A), [Bibek Bhusal](https://arxiv.org/search/cs?searchtype=author&query=Bhusal,+B), [Danning Di](https://arxiv.org/search/cs?searchtype=author&query=Di,+D), [Yuan Jin](https://arxiv.org/search/cs?searchtype=author&query=Jin,+Y), [Meet Raval](https://arxiv.org/search/cs?searchtype=author&query=Raval,+M), [Zhiwen Chen](https://arxiv.org/search/cs?searchtype=author&query=Chen,+Z), [Boyao Sun](https://arxiv.org/search/cs?searchtype=author&query=Sun,+B), [Shuguang Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+S), [Yunlong He](https://arxiv.org/search/cs?searchtype=author&query=He,+Y), [Yantao Yao](https://arxiv.org/search/cs?searchtype=author&query=Yao,+Y), [Sagar Chordia](https://arxiv.org/search/cs?searchtype=author&query=Chordia,+S), [Wenlin Chen](https://arxiv.org/search/cs?searchtype=author&query=Chen,+W), [Santanu Kolay](https://arxiv.org/search/cs?searchtype=author&query=Kolay,+S), [Qin Huang](https://arxiv.org/search/cs?searchtype=author&query=Huang,+Q), [Ellie Wen](https://arxiv.org/search/cs?searchtype=author&query=Wen,+E)

[View PDF](https://arxiv.org/pdf/2608.10562) [HTML (experimental)](https://arxiv.org/html/2608.10562v1)

> Abstract:Not all clicks are equal. Industrial ads ranking decouples conversion probability into click-through rate (CTR) and post-click conversion rate (CVR), yet treats every click as the same event. In reality, users provide a free, self-generated signal of intent through their physical UI interactions. Different click types on the same ad exhibit a 4-fold difference in actual conversion rates. By conflating these signals, the standard CVR model under-predicts high-intent clicks and over-predicts low-intent ones, which is a bias masked by near-perfect aggregate calibration. We propose MARCO (Multi-intent Ads Ranking Composition Optimization), a framework that resolves this bias by decomposing each click by intent. Using the logged click type as a free behavioral label, MARCO trains per-intent CVR heads on homogeneous populations, and at serving time composes their per-intent CVR estimates under a predicted distribution over intents. Theoretically, we prove that decomposition never raises population risk, give the exact headroom under squared loss and non-negativity under the deployed loss, and show through a routing-efficiency dial how much of it reaches serving. Because the population-optimal score is unchanged, any gain is a finite-capacity estimation and calibration effect that we validated both offline and online. For deployment at scale, we further cast multi-impression, multi-click attribution as credit assignment with a bias-variance tradeoff analogous to RL return estimation, showing last-impression, first-click attribution is the low-bias, low-variance, deterministic choice under production constraints, and derive three consistency conditions enforced end-to-end at scale. Deployed at binary intent granularity, MARCO corrects per-intent calibration to approximately 100%, lifts conversions per click by +2.80%, and drives +0.98% cumulative improvement in topline metrics.

| Subjects: | Machine Learning (cs.LG) |
| --- | --- |
| Cite as: | [arXiv:2608.10562](https://arxiv.org/abs/2608.10562) \[cs.LG\] |
|  | (or [arXiv:2608.10562v1](https://arxiv.org/abs/2608.10562v1) \[cs.LG\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.10562](https://doi.org/10.48550/arXiv.2608.10562) |

## Submission history

From: Xiru Huang \[[view email](https://arxiv.org/show-email/adab314f/2608.10562)\]  
**\[v1\]** Tue, 11 Aug 2026 06:49:59 UTC (631 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.10562) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))