Pith. sign in

REVIEW 2 major objections 1 minor 26 references

HMAF uses a Plan-Calibrate-Execute structure to allocate impressions between guaranteed delivery contracts and real-time bidding across multiple slots.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

HMAF is a Plan-Calibrate-Execute framework for GD-RTB impression allocation that claims to improve delivery rates and revenue when deployed at Meituan.

T0 review reviewed 2026-06-27 challenge →

load-bearing objection HMAF describes a Plan-Calibrate-Execute framework for multi-slot GD-RTB allocation and reports deployment lifts at Meituan, but the abstract gives no experimental controls or comparisons to support the causal claim. the 2 major comments →

arxiv 2606.09896 v1 pith:IX6PNM53 submitted 2026-06-04 cs.GT cs.AIcs.LG

HMAF: A Hierarchical Multi-Slot GD-RTB Allocation Framework

Tianxing Bu , Zhaoqi Zhang , Linyou Cai , Miao Xie , Shengri Xue , Tan Qu , Qianlong Xie , Xingxing Wang
show 2 more authors
Siqiang Luo Gao Cong
This is my paper
classification cs.GT cs.AIcs.LG
keywords guaranteed deliveryreal-time biddingimpression allocationhierarchical frameworkmulti-slot optimizationonline advertising
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper introduces HMAF as a unified framework for optimizing impression allocation when guaranteed delivery contracts coexist with real-time bidding auctions. It structures the process around offline constraint optimization, dynamic competitiveness calibration, and real-time listwise ranking to handle multi-slot constraints and tradeoffs between revenue and contract fulfillment. Prior methods that decouple the two or apply heuristic priorities fall short in complex environments. Deployment in live marketing scenarios produced measured lifts in delivery rate and total revenue.

Core claim

HMAF employs the Plan-Calibrate-Execute paradigm as its core structure, and integrates offline constraint optimization with online decision-making, balancing offline GD resource planning, dynamically calibrating GD-RTB competitiveness, and making real-time listwise rank decisions across multi-slot environments.

What carries the argument

The Plan-Calibrate-Execute paradigm, which links offline GD resource planning, online GD-RTB competitiveness calibration, and real-time listwise ranking to manage multi-slot constraints.

Load-bearing premise

The Plan-Calibrate-Execute paradigm can effectively balance offline GD resource planning with online GD-RTB competitiveness calibration without introducing new conflicts or inefficiencies in multi-slot environments.

What would settle it

A live A/B test on an advertising platform that compares HMAF against decoupled or heuristic baselines and finds no lift in GD delivery rate or total revenue.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Effective coordination of long-term contract delivery with short-term revenue maximization under multi-slot impression limits.
  • Reduced reliance on heuristic priority rules that fail to balance GD and RTB objectives.
  • Scalable real-time decisions that respect both offline plans and dynamic auction competitiveness.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The hierarchical separation of planning and calibration stages may apply to other hybrid contract-auction resource systems.
  • It could reduce inefficiencies that arise when priority rules dominate allocation in competitive online marketplaces.
  • Testing the framework on platforms with different slot counts or contract densities would clarify its robustness beyond the reported deployment.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 1 minor

Summary. The paper proposes HMAF, a hierarchical multi-slot allocation framework for coexisting GD contracts and RTB auctions in online advertising. It introduces a Plan-Calibrate-Execute paradigm that combines offline constraint optimization for GD resource planning, online competitiveness calibration between GD and RTB, and real-time listwise ranking decisions. The central claim is that this unified approach outperforms decoupled or heuristic methods and, when deployed at Meituan, produced a 3.72% lift in GD delivery rate and a 1.59% increase in total ad revenue.

Significance. If the reported deployment gains can be causally attributed to the proposed components via controlled experiments, the work would offer a practical contribution to industrial ad allocation systems by addressing the tension between long-term contract delivery and short-term revenue in multi-slot settings. The integration of offline planning with online calibration is a relevant direction for platforms where GD and RTB compete for the same impressions.

major comments (2)
  1. [Abstract] Abstract (and any results section): The headline empirical claim of 3.72% GD delivery rate improvement and 1.59% revenue increase is presented without any description of the A/B test protocol, control-group definition, pre/post statistical tests, ablation against the prior system, or controls for confounding platform changes. This attribution step is load-bearing for the central claim that HMAF is responsible for the observed deltas.
  2. The Plan-Calibrate-Execute paradigm is described at a high level but no concrete formulation, objective functions, or constraint sets are provided for the offline planning stage or the online calibration stage, preventing assessment of whether the framework actually resolves the multi-slot impression constraints mentioned in the abstract.
minor comments (1)
  1. Notation for multi-slot ranking and competitiveness calibration is introduced without explicit definitions or pseudocode, making the online decision-making component difficult to follow.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. We address each major comment below. Where the comments identify gaps in experimental description or mathematical detail, we agree that revisions are warranted and will incorporate them.

read point-by-point responses

  1. Referee: [Abstract] Abstract (and any results section): The headline empirical claim of 3.72% GD delivery rate improvement and 1.59% revenue increase is presented without any description of the A/B test protocol, control-group definition, pre/post statistical tests, ablation against the prior system, or controls for confounding platform changes. This attribution step is load-bearing for the central claim that HMAF is responsible for the observed deltas.

    Authors: We agree that the current presentation of the deployment results is insufficiently detailed. In the revised manuscript we will expand the results section (and update the abstract accordingly) to describe the A/B test protocol, the definition of the control group (the prior production system), test duration, and the statistical tests used. We will also clarify that the reported lifts are relative to the immediately preceding system and note any platform-wide changes that were controlled for during the evaluation window. Some internal operational metrics will necessarily remain summarized for confidentiality reasons. revision: partial

  2. Referee: [—] The Plan-Calibrate-Execute paradigm is described at a high level but no concrete formulation, objective functions, or constraint sets are provided for the offline planning stage or the online calibration stage, preventing assessment of whether the framework actually resolves the multi-slot impression constraints mentioned in the abstract.

    Authors: The referee correctly observes that the manuscript currently presents the paradigm at a conceptual level. In the revision we will move the explicit mathematical formulations—specifically the offline multi-slot constraint optimization problem (objective and constraint sets) and the online competitiveness calibration objective—into the main body of Sections 3 and 4, with the full constraint sets and decision variables stated. This will allow readers to verify how the framework addresses the multi-slot impression constraints. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper describes an applied allocation framework (HMAF) built around the Plan-Calibrate-Execute structure and reports deployment lifts at Meituan as external outcomes. No equations, fitted-parameter predictions, self-citations, or uniqueness theorems appear in the supplied text that would reduce any claimed result to its own inputs by construction. The empirical percentages are presented as measured implementation effects rather than derived quantities, leaving the argument self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Insufficient information in the abstract to identify specific free parameters, axioms, or invented entities.

reviewed 2026-06-27 · how reviews work

0 comments
Cite this review

Pith. "Pith review of HMAF: A Hierarchical Multi-Slot GD-RTB Allocation Framework." pith.science (2026). https://pith.science/paper/IX6PNM53

@misc{pith2026260609896,
  author       = {Pith},
  title        = {Pith review of: HMAF: A Hierarchical Multi-Slot GD-RTB Allocation Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IX6PNM53}},
  note         = {Machine review of arXiv:2606.09896}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

In modern online advertising platforms, Guaranteed Delivery (GD) contracts coexist and bid with Real-Time Bidding (RTB) auctions. Recent approaches either decouple GD and RTB optimization or rely on heuristic priority rules, and thus fail to effectively balance short-term revenue maximization with long-term contract delivery under complex multi-slot delivery and impression constraints. To address these challenges, we propose HMAF (Hierarchical Multi-Slot Allocation Framework), a unified framework designed to optimize impression allocation in GD--RTB advertising platforms. HMAF employs the Plan--Calibrate--Execute paradigm as its core structure, and integrates offline constraint optimization with online decision-making, balancing offline GD resource planning, dynamically calibrating GD--RTB competitiveness, and making real-time listwise rank decisions across multi-slot environments. HMAF has been implemented in multiple marketing scenarios at Meituan, one of the world's largest online food delivery platforms, leading to a 3.72% increase in GD delivery rate and a 1.59% increase in total advertisement revenue.

Figures

Figures reproduced from arXiv: 2606.09896 by Gao Cong, Linyou Cai, Miao Xie, Qianlong Xie, Shengri Xue, Siqiang Luo, Tan Qu, Tianxing Bu, Xingxing Wang, Zhaoqi Zhang.

Figure 1
Figure 1. Figure 1: Flowchart of Meituan display advertising system [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of HMAF. 2.2 Guaranteed Delivery Optimization In addition to allocation, several works focus on optimizing the ful￾fillment of guaranteed delivery ads. Wu et al. [19] explore auction￾based impression allocation, using multi-agent reinforcement learn￾ing to optimize allocation decisions and adapt to dynamic traffic patterns. Zhang et al. [21] propose a control-based bidding approach for mobile live… view at source ↗
Figure 3
Figure 3. Figure 3: Parameter sensitivity analysis of HMAF. competitiveness coefficient, influencing the allocation of resources between GD and RTB ads;𝜓1 and𝜓2 control the relative importance of the loss terms, with 𝜓1 balancing the importance of the page￾view constrained allocation and𝜓2 managing the trade-off between real-time revenue and long-term contract fulfillment. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 1 canonical work pages · 1 internal anchor

  1. [1]

    Anand Bhalgat, Jon Feldman, and Vahab Mirrokni. 2012. Online allocation of display ads with smooth delivery. InProceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining. 1213–1221

    2012

  2. [2]

    Vijay Bharadwaj, Peiji Chen, Wenjing Ma, Chandrashekhar Nagarajan, John Tomlin, Sergei Vassilvitskii, Erik Vee, and Jian Yang. 2012. Shale: an efficient algorithm for allocation of guaranteed display advertising. InProceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining. 1195–1203

    2012

  3. [3]

    Ye Chen, Pavel Berkhin, Bo Anderson, and Nikhil R Devanur. 2011. Real-time bidding algorithms for performance-based display ad allocation. InProceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining. 1307–1315

    2011

  4. [4]

    Xiao Cheng, Chuanren Liu, Liang Dai, Peng Zhang, Zhen Fang, and Zhonglin Zu. 2022. An adaptive unified allocation framework for guaranteed display advertising. InProceedings of the Fifteenth ACM International Conference on Web Search and Data Mining. 132–140

    2022

  5. [5]

    Liang Dai, Zhonglin Zu, Hao Wu, Liang Wang, and Bo Zheng. 2023. Fairness- aware guaranteed display advertising allocation under traffic cost constraint. In Proceedings of the ACM Web Conference 2023. 3572–3580

    2023

  6. [6]

    Zhen Fang, Yang Li, Chuanren Liu, Wenxiang Zhu, Yu Zheng, and Wenjun Zhou

  7. [7]

    In2019 IEEE International Conference on Data Mining (ICDM)

    Large-scale personalized delivery for guaranteed display advertising with real-time pacing. In2019 IEEE International Conference on Data Mining (ICDM). IEEE, 190–199. KDD ’26, August 09–13, 2026, Jeju Island, Republic of Korea Tianxing Bu et al

    2026

  8. [8]

    Arpita Ghosh, Preston McAfee, Kishore Papineni, and Sergei Vassilvitskii. 2009. Bidding for representative allocations for display advertising. InInternational workshop on internet and network economics. Springer, 208–219

    2009

  9. [9]

    Xiang He, Wuyang Mao, Zhenghang Xu, Yuanzhe Gu, Yundu Huang, Zhonglin Zu, Liang Wang, Mengyu Zhao, and Mengchuan Zou. 2024. An Efficient Local Search Algorithm for Large GD Advertising Inventory Allocation with Multilinear Constraints. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1040–1049

    2024

  10. [10]

    Ali Hojjat, John Turner, Suleyman Cetintas, and Jian Yang. 2014. Delivering guaranteed display ads under reach and frequency requirements. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 28

    2014

  11. [11]

    2004.Knapsack Problems

    Hans Kellerer, Ulrich Pferschy, and David Pisinger. 2004.Knapsack Problems. Springer, Berlin, Heidelberg

    2004

  12. [12]

    Yan Li, Yundu Huang, Wuyang Mao, Furong Ye, Xiang He, Zhonglin Zu, and Shaowei Cai. 2024. Bi-Objective Contract Allocation for Guaranteed Delivery Advertising. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1691–1700

    2024

  13. [13]

    Wuyang Mao, Chuanren Liu, Yundu Huang, Zhonglin Zu, M Harshvardhan, Liang Wang, and Bo Zheng. 2023. End-to-end inventory prediction and contract allocation for guaranteed delivery advertising. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1677–1686

    2023

  14. [14]

    1990.Knapsack problems: algorithms and com- puter implementations

    Silvano Martello and Paolo Toth. 1990.Knapsack problems: algorithms and com- puter implementations. John Wiley & Sons, Inc

    1990

  15. [15]

    Erik Vee, Sergei Vassilvitskii, and Jayavel Shanmugasundaram. 2010. Optimal online assignment with forecasts. InProceedings of the 11th ACM conference on Electronic commerce. 109–118

    2010

  16. [16]

    Jun Wang and Shuai Yuan. 2015. Real-time bidding: A new frontier of com- putational advertising research. InProceedings of the Eighth ACM International Conference on Web Search and Data Mining. 415–416

    2015

  17. [17]

    Xiaoyu Wang, Yonghui Guo, Bin Tan, Tao Yang, Dongbo Huang, Lan Xu, Hao Zhou, and Xiangyang Li. 2024. Follow the libra: guiding fair policy for unified impression allocation via adversarial rewarding. InProceedings of the 17th ACM International Conference on Web Search and Data Mining. 750–759

    2024

  18. [18]

    XiaoYu Wang, Bin Tan, Yonghui Guo, Tao Yang, Dongbo Huang, Lan Xu, Niko- laos M Freris, Hao Zhou, and Xiang-Yang Li. 2022. CONFLUX: A Request-level Fusion Framework for Impression Allocation via Cascade Distillation. InPro- ceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 4070–4078

    2022

  19. [19]

    Penghui Wei, Yongqiang Chen, Shaoguo Liu, Liang Wang, and Bo Zheng. 2023. RLTP: Reinforcement Learning to Pace for Delayed Impression Modeling in Preloaded Ads. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 5204–5214

    2023

  20. [20]

    Di Wu, Cheng Chen, Xiujun Chen, Junwei Pan, Xun Yang, Qing Tan, Jian Xu, and Kuang-Chih Lee. 2021. Impression allocation and policy search in display advertising. In2021 IEEE International Conference on Data Mining (ICDM). IEEE, 749–756

    2021

  21. [21]

    Shuai Yuan, Jun Wang, and Xiaoxue Zhao. 2013. Real-time bidding for online advertising: measurement and analysis. InProceedings of the seventh international workshop on data mining for online advertising. 1–8

    2013

  22. [22]

    Haoqi Zhang, Junqi Jin, Zhenzhe Zheng, Fan Wu, Haiyang Xu, and Jian Xu. 2022. Control-based Bidding for Mobile Livestreaming Ads with Exposure Guarantee. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management. 2539–2548

    2022

  23. [23]

    Hong Zhang, Lan Zhang, Lan Xu, Xiaoyang Ma, Zhengtao Wu, Cong Tang, Wei Xu, and Yiguo Yang. 2020. A request-level guaranteed delivery advertising planning: Forecasting and allocation. InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2980–2988

    2020

  24. [24]

    Weinan Zhang, Yifei Rong, Jun Wang, Tianchi Zhu, and Xiaofan Wang. 2016. Feedback control of real-time display advertising. InProceedings of the Ninth ACM International Conference on Web Search and Data Mining. 407–416

    2016

  25. [25]

    Weinan Zhang, Shuai Yuan, and Jun Wang. 2014. Optimal real-time bidding for display advertising. InProceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. 1077–1086

    2014

  26. [26]

    Zhaoqi Zhang, Jiaming Deng, Miao Xie, Linyou Cai, Qianlong Xie, Xingxing Wang, Siqiang Luo, and Gao Cong. 2026. Beyond Single Slot: Joint Optimization for Multi-Slot Guaranteed Display Advertising.arXiv preprint arXiv:2605.21556 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

This paper was first reviewed by grok-4.3 on June 27, 2026.