Pith. sign in

REVIEW 5 major objections 5 minor 33 references

IC Mechanisms for Risk-Averse Advertisers in the Online Advertising System

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read DFP adjusts per-click payments using real-time conversion data to keep advertisers' realized CPA equal to their target, which the authors claim preserves truthful bidding even with sparse clicks.

desk verdict TIC is a genuinely new concept and the CFP result is clean, but the main theorem that DFP is TIC is circular and only holds for a clairvoyant payment rule; the online variant is unsupported. read the letter →

arxiv 2411.13162 v1 pith:LQUOMQ72 submitted 2024-11-20 cs.GT

Bingzhe Wang , Ruohan Qian , Yuejia Dou , Qi Qi , Bo Shen , Changyuan Li , Yixuan Zhang , Yixin Su
show 7 more authors
Xin Yuan Wenqiang liu Bin Zou Wen Yi Zhi Guo Shuanglong Li Liu Lin
This is my paper · ORCID
classification cs.GT
keywords advertisersmechanismmechanismsadvertisingautobiddingduringexpectationsoutcomes
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

In online advertising, many systems let advertisers report a target cost per acquisition (tCPA), then the platform bids on their behalf and charges them per click. In an autobidding incentive compatible mechanism, an advertiser's best strategy is to report her true tCPA, but only in expectation. In practice, an advertiser with few clicks may see her realized cost per acquisition drift above the target, and if she is risk averse she may change bids or leave the system.

The paper defines a stronger property, Time-Invariant Incentive Compatibility (TIC): truthful bidding should remain optimal during the auction process, stage by stage, not just in expectation. It proposes Coupled First-Price Auction (CFP) as a baseline AIC mechanism, then Decoupled First-Price Auction (DFP), where allocation and payment are designed separately. In DFP, the platform observes real-time clicks and conversions, and chooses per-click payments so that the realized CPA equals the advertiser's reported tCPA in every stage. A PPO-based reinforcement learning controller is used to make these payment decisions online, because the true conversion count for a stage is not known when the clicks happen.

The key theoretical result is that DFP is TIC. The catch is that this result is built into the design: the payment optimization exactly targets the condition that defines TIC. In the online system, the controller must estimate future conversions, and the paper provides no guarantee that TIC survives estimation error. The experiments report that DFP keeps tCPA/CPA close to 1, but this is the same quantity the RL reward is trained to minimize.

Extended reading notes

Core claim

DFP is TIC (Theorem 2): in the OCPC model with tCPA bidders, the Decoupled First-Price Auction is Time-Invariant Incentive Compatible, meaning truthful bidding is the optimal strategy for risk-averse bidders during the auction process, in every stage, even with sparse clicks.

Load-bearing premise

The payment mechanism bPPl_t chooses p* using bZmNt, the realized conversion count for the entire stage, but per-click payments in the real system must be set at click time before delayed conversion feedback arrives. The theorem therefore only holds under the assumption that the platform knows the stage's future conversions (or that the RL estimate is exact enough to preserve the equality), and the paper provides no bound or guarantee for the estimated case.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper studies autobidding for tCPA bidders under the OCPC pricing model. It first proposes Coupled First-Price Auction (CFP), a first-price mechanism claimed to be Autobidding Incentive Compatible (AIC). It then introduces Time-Invariant Incentive Compatibility (TIC), a stricter notion intended to capture the behavior of risk-averse bidders who evaluate realized outcomes at the end of each stage, and proposes Decoupled First-Price Auction (DFP), in which the payment rule is defined as the solution of the stage-level optimization problem bPPl_t. A PPO-based RL algorithm with a two-part reward is proposed to implement the DFP payment rule. Experiments on 31 days of data from 5000 bidders compare tCPA/CPA ratios and payment fluctuation across CPA, Pacing, CFP, and DFP.

Significance. If DFP were truly incentive compatible in a causal online sense, the paper would address a real practical problem: sparse-click advertisers deviating from truthful bidding because realized CPAs deviate from the tCPA. The first-price AIC observation for CFP is clean under the paper's own definitions, and the empirical comparison over 155,000 bidder-day instances is a useful descriptive exercise. However, the central theoretical claim is not supported as stated: Theorem 2 holds only for a hindsight payment rule that uses realized stage-level conversions and clicks, and the TIC concept is built on a behavioral rule that presupposes the desired equivalence. The paper provides no formal model of risk-averse utility maximization, no approximation guarantee for the RL-based causal payment rule, and no code or machine-checked proofs. The empirical results are suggestive but do not repair the gap between the deployed mechanism and the theorem.

major comments (5)
  1. [§4.2, Definition 7 and bPPl_t] The proof of Theorem 2 assumes the platform knows the stage-level realized conversions bZmNt and clicks bymnk when choosing the per-click payments bpmnk. The objective of bPPl_t is sum_m | bZmNt * tCPA_m / (sum_{n,k} bymnk * bpmnk) - 1 |, and the optimal p* is defined after these whole-stage realized values are available. In OCPC, per-click payments must be committed at click time, before delayed conversions (and often before later clicks) are observed; Section 4.2 itself acknowledges that "the platform lacks knowledge of bZmNt when determining bpmnk." Theorem 2 therefore proves TIC only for a clairvoyant hindsight mechanism, not for the causal payment rule that is actually deployed.
  2. [§4.2, Lemma 6 and Theorem 2] Lemma 6 is essentially a restatement of Lemma 4: the condition bZmNt * tCPA_m = sum bymnk * bpmnk is exactly the requirement tCPA_m / CPA_mNt = 1. It is not specific to CFP, and the bPPl_t objective is deliberately designed to make this equality hold. Substituting p* into the objective and concluding that the objective value is zero is true by construction, so Theorem 2 has no independent game-theoretic content as a statement about bidder incentives.
  3. [§5.1, reward r^(1)_n and state definition] The RL implementation does not close the oracle gap. The reward r^(1)_n is defined using the whole-stage realized quantities bZmNt and bymnk, and the state description states that bZmn is updated only once at the end of each stage. The empirical tCPA/CPA ratios in Table 1 and Figures 3-5 therefore mainly show how well the reward function is optimized, not that truthful bidding is an optimal strategy in a mechanism whose payments are set causally at click time. No theorem or bound is given for the error introduced by estimating bZmNt, so the main claim is not established for the deployed variant.
  4. [Definition 5 and Lemma 3] The paper models risk aversion as a behavioral rule rather than as a preference. Definition 5 says a bidder is risk-averse if she adjusts her bids or withdraws when she observes deviations, and Lemma 3 then asserts that such a bidder bids truthfully iff tCPA_m / CPA_mNt = 1 at every stage. This is not derived from any utility maximization problem; it is an assumption that the desired equivalence holds. Consequently, TIC (Definition 6) is a post-hoc consistency condition on realized CPA rather than an incentive property established for risk-averse utility maximizers.
  5. [Appendix C.6, proof of Lemma 5] The derivation of the click-volume threshold in Lemma 5 controls the two Chernoff tails separately and then takes the larger threshold, but for the conjunction of both inequalities to hold with probability at least 1 - epsilon, a union bound would require each tail to be controlled at about epsilon/2 rather than epsilon. Moreover, Chernoff gives only an upper bound on tail probabilities, so it cannot establish the stated "necessary" condition on bYmNt; at best the derivation gives a sufficient condition under the specific concentration bound used. The lemma as stated is therefore not rigorous.
minor comments (5)
  1. [Throughout] There are multiple typos and rendering problems: "desgin" in Section 1, "vloume" in Section 3, "MES" for MSE in Section 5.2, "clcik" in Figure 2, and "an bidder" in Section 6.2.2 and Appendix A. The accents used for expected and realized values, e.g. [CPAmNt and bPmn, are also not rendered consistently, which makes the distinction between ex ante and realized objects hard to follow.
  2. [§4.2, bPPl_t] The text says that if bZmNt were known, bPPl_t would reduce to a standard online decision-making problem, but the objective is evaluated on whole-stage realized clicks and conversions; this is a hindsight problem, not an online one. The wording should be corrected to avoid implying causal information is available.
  3. [§6.2.4 and Table 1] The payment fluctuation comparison mixes different information regimes: CPA and Pacing are offline baselines with complete knowledge of all outcomes, while DFP is the deployed system. The descriptive quartile comparison is useful, but the text should acknowledge that this is not a controlled online experiment and should report the number of clicks or impressions underlying the variance statistics.
  4. [§5.2] The PPO, GAE, and entropy-regularization components are described without citations; standard references for proximal policy optimization and generalized advantage estimation should be added, since the paper claims to use these specific methods.
  5. [§4.2 and Definition 7] The case of a stage with zero clicks or zero conversions is not handled: the ratio in bPPl_t is undefined when the denominator is zero, and Lemma 4 similarly assumes CPA_mNt is well defined. Given that the paper emphasizes sparse-click systems, this boundary case should be addressed explicitly.

Circularity Check

2 steps flagged · score 8.0 of 10

DFP's TIC property is installed by definition: the payment optimization bPPl_t minimizes exactly the TIC deviation, and the RL/experimental metric is the same ratio.

  1. self definitional [Definition 7, Section 4.2; Lemma 6; Theorem 2]
    "min sum_{m in [M]} | (bZmNt * tCPA_m) / (sum_{n in [Nt], k in [K]} bymnk * bpmnk) - 1 | (bPPl_t) ... Payment Mechanism p: bpmnk = p*_mnk, where p*_mnk is the optimal solution to bPPl_t. By bpmnk = p*_mnk, we have sum_{m in [M]} | ... - 1 | = 0. Then, DFP is TIC."

    Lemma 6 defines TIC for CFP as bZmNt * tCPA_m = sum bymnk * bpmnk, i.e. the bracketed ratio is exactly 1. The optimization bPPl_t is the minimization of the absolute deviation of that very ratio, and Definition 7 sets the payment to p*, the optimizer of bPPl_t. Therefore the equality required by the TIC condition is imposed by construction, not derived. Theorem 2's proof simply substitutes p* into the objective and reads off zero; it proves that DFP is TIC because DFP's payment rule was defined to make the TIC condition true.

  2. fitted input called prediction [Section 5.1 (reward function) and Section 6.2.1 (TIC evaluation)]
    "r(1)_n = - log sum_{m in [M]} | (sum_{n in [Nt],k in [K]} bymnk * bpmnk) / (bZmNt * tCPA_m + xi) - 1 | ... Drawing upon Lemma 4, we quantify TIC using the ratio tCPA/CPA."

    The RL reward r(1) is a monotone transformation of the same absolute deviation that bPPl_t minimizes, and the experimental TIC metric tCPA/CPA is exactly the ratio appearing inside that deviation. Thus the empirical claim that DFP is closest to TIC reports the objective that the PPO algorithm was trained to minimize, rather than an independent, out-of-sample test of the mechanism's incentive properties. The measurement is therefore statistically forced by the choice of reward function.

full rationale

The central result, Theorem 2, is circular in the strong sense: TIC is characterized by Lemma 4 as the condition tCPA/CPA = 1 per stage, Lemma 6 rewrites this as an equality involving realized conversions and payments, and the DFP payment rule is defined as the optimal solution of an optimization problem whose objective is the absolute deviation from that equality. Substituting the optimizer into the objective gives zero by construction, so the proof of Theorem 2 is a one-line restatement of the definition of the payment rule rather than an independent derivation. The paper does not appear to rely on load-bearing self-citations; the circularity is internal to the mechanism definition. A further, non-circular but serious limitation is that bPPl_t and the RL reward use the stage-realized conversion count bZmNt, which the platform admits it does not know when setting per-click payments, so the theorem characterizes a clairvoyant payment rule rather than the causally implementable online mechanism. The empirical section compounds the issue by measuring the same ratio that the reward function minimizes. For these reasons, the paper's headline TIC claim reduces by definition to its own payment-optimization input.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claim relies on a behavioral model of risk aversion, on common-knowledge CTR/CVR, and on the unstated feasibility of setting payments after observing stage-level conversions. The TIC condition itself is both the target and the optimization objective.

free parameters (5)
  • epsilon (epsilon)
    Tolerance in epsilon-TIC and epsilon-risk-averse definitions; no value is calibrated, yet Section 6.2.3 reports epsilon_DFP <= 0.1.
  • zeta (zeta)
    Weight between the TIC reward r1 and the payment-fluctuation reward r2 in Section 5.1; no value is reported.
  • xi (xi)
    Constant added to the denominator of r1 to prevent division by zero; no value is reported.
  • PPO hyperparameters (kappa, gamma, lambda, alpha1, alpha2, alpha3)
    Clip range, discount factors, GAE lambda, and loss weights in Section 5.2; no values or tuning details are given.
  • stage partition T = 31 days in the main experiment
    The risk-averse bidder's stage length is exogenous; the experiments set one stage per day and also test CFP-tau with 2 to 8 days.
assumptions (6)
  • domain assumption CTR and CVR are common knowledge and precisely estimated for every bidder, round, and slot.
    Stated in Section 3: both rates are considered common knowledge shared between platform and bidder, and precisely estimated. All mechanism proofs use these rates as fixed inputs.
  • domain assumption Bidders are value-maximizers subject to a hard tCPA constraint.
    Definition 1 and problem P_m define utility as total conversion value with CPA <= tCPA; the incentive analysis only applies to this bidder type.
  • ad hoc to paper A risk-averse bidder bids truthfully exactly when the realized CPA equals tCPA at every stage.
    Definition 5 equates risk aversion with adjusting bids when actual outcomes deviate from expectations, and Lemma 3 converts this into the ratio condition tCPA / [CPA] = 1. This is a behavioral postulate, not derived from a utility function.
  • domain assumption Clicks and conversions are independent Bernoulli variables with known rates, and the actual payment per click equals the expected payment given a click.
    Used in Lemma 5's Chernoff bound; the model defines binary click and conversion outcomes and treats the CFP payment bm * cvr as deterministic once a click occurs.
  • ad hoc to paper For each stage, there exist nonnegative per-click payments that exactly satisfy Zhat * tCPA = sum(phat * yhat) for every bidder.
    Definition 7 assumes the minimizer p* of bPPl_t achieves objective value zero, which requires enough clicks and unrestricted payment levels; no feasibility condition is stated.
  • ad hoc to paper The platform can set per-stage payments using the stage's realized conversion count bZmNt.
    bPPl_t and Theorem 2 use bZmNt as an input to the payment decision, but payments must be set per click before delayed conversion feedback arrives. The paper acknowledges this gap and proposes RL estimation without a guarantee.

how reviews work

0 comments
Cite this review

Pith. "Pith review of IC Mechanisms for Risk-Averse Advertisers in the Online Advertising System." pith.science (2026). https://pith.science/paper/LQUOMQ72

@misc{pith2026241113162,
  author       = {Pith},
  title        = {Pith review of: IC Mechanisms for Risk-Averse Advertisers in the Online Advertising System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LQUOMQ72}},
  note         = {Machine review of arXiv:2411.13162}
}
read the original abstract

The autobidding system generates huge revenue for advertising platforms, garnering substantial research attention. Existing studies in autobidding systems focus on designing Autobidding Incentive Compatible (AIC) mechanisms, where the mechanism is Incentive Compatible (IC) under ex ante expectations. However, upon deploying AIC mechanisms in advertising platforms, we observe a notable deviation between the actual auction outcomes and these expectations during runtime, particularly in the scene with few clicks (sparse-click). This discrepancy undermines truthful bidding among advertisers in AIC mechanisms, especially for risk-averse advertisers who are averse to outcomes that do not align with the expectations. To address this issue, we propose a mechanism, Decoupled First-Price Auction (DFP), that retains its IC property even during runtime. DFP dynamically adjusts the payment based on real-time user conversion outcomes, ensuring that advertisers' realized utilities closely approximate their expected utilities during runtime. To realize the payment mechanism of DFP, we propose a PPO-based RL algorithm, with a meticulously crafted reward function. This algorithm dynamically adjusts the payment to fit DFP mechanism. We conduct extensive experiments leveraging real-world data to validate our findings.

Figures

Figures reproduced from arXiv: 2411.13162 by the authors.

Figure 1
Figure 1. The Conversion Volume of an Advertiser operate independently. In this paper, we demonstrate the theoretical feasibility of decoupling the intricate problem of mechanism design into two distinct, yet interdependent, online optimization problems, aiming to design allocation mechanisms and payment mechanisms, respectively. The allocation mechanism is to maximize certain objectives of the platform. We insert real-time o… view at source ↗
Figure 2
Figure 2. Decoupled First-Price Auction The allocation mechanism of DFP is a canonical online optimization problem, which can be addressed utilizing various algorithms such as online algorithms, deep learning, and reinforcement learning (RL), among others [17, 18]. As we can leverage any allocation mechanism in DFP, it is not the focus of this paper and we focus on the payment mechanism design. Distinct from the majority of o… view at source ↗
Figure 3
Figure 3. The tCP A CP A ratio of 5000 bidders over 31 days. The experimental results demonstrate that CPA is TIC, since CPA operates under an offline setting, representing the optimal solution among all online mechanisms. The remaining three mechanisms are ϵ-TIC, with DFP exhibiting the closest ratio of tCP A CP A to 1, thus incurring the smallest ϵDF P . This empirical evidence underscores that in the online setting, DFP is… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison between CFP and DFP. 11 [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: The tCP A CP A ratio of a bidder with an average daily click volume of 132. Notably, even though the bidder’s average daily click volume is significantly below the threshold stipulated in Lemma 5, the condition ϵDF P ≤ 0.1 held almost invariably, with ϵDF P notably sma…
Figure 6
Figure 6. Figure 6: The tCP A CP A ratio of a bidder with an average daily click volume of 1308. Notably, even though the bidder’s average daily click volume is significantly below the threshold stipulated in Lemma 5, the condition ϵDF P ≤ 0.1 held almost invariably, with ϵDF P notably sm…
Figure 7
Figure 7. Figure 7: Per-click payments of the sampled bidder over 31 days. [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 26 canonical work pages

  1. [1]

    Counterspeculation, auctions, and competitive sealed tenders.The Journal of finance, 16(1):8–37, 1961

    William Vickrey. Counterspeculation, auctions, and competitive sealed tenders.The Journal of finance, 16(1):8–37, 1961

    1961

  2. [2]

    Multipart pricing of public goods

    Edward H Clarke. Multipart pricing of public goods. Public choice, pages 17–33, 1971

    1971

  3. [3]

    Incentives in teams

    Theodore Groves. Incentives in teams. Econometrica: Journal of the Econometric Society, pages 617–631, 1973

    1973

  4. [4]

    Optimal auction design

    Roger B Myerson. Optimal auction design. Mathematics of operations research, 6(1):58–73, 1981

    1981

  5. [5]

    Greedy bidding strategies for keyword auctions

    Matthew Cary, Aparna Das, Ben Edelman, Ioannis Giotis, Kurtis Heimerl, Anna R Karlin, Claire Mathieu, and Michael Schwarz. Greedy bidding strategies for keyword auctions. In Proceedings of the 8th ACM Conference on Electronic Commerce, pages 262–271, 2007

    work page 2007

  6. [6]

    Sponsored search auctions

    Sébastien Lahaie, David M Pennock, Amin Saberi, and Rakesh V V ohra. Sponsored search auctions. Algorithmic game theory, 1:699–716, 2007

    work page 2007

  7. [7]

    Strategic bidder behavior in sponsored search auctions

    Benjamin Edelman and Michael Ostrovsky. Strategic bidder behavior in sponsored search auctions. Decision support systems, 43(1):192–198, 2007

    work page 2007

  8. [8]

    Sponsored search: an overview of the concept, history, and technology

    Bernard J Jansen and Tracy Mullen. Sponsored search: an overview of the concept, history, and technology. International Journal of Electronic Business, 6(2):114–131, 2008

    work page 2008

Show all 33 references
  1. [9]

    Internet advertising and the generalized second- price auction: Selling billions of dollars worth of keywords

    Benjamin Edelman, Michael Ostrovsky, and Michael Schwarz. Internet advertising and the generalized second- price auction: Selling billions of dollars worth of keywords. American economic review, 97(1):242–259, 2007

    2007

  2. [10]

    Autobidding with constraints

    Gagan Aggarwal, Ashwinkumar Badanidiyuru, and Aranyak Mehta. Autobidding with constraints. In Web and Internet Economics: 15th International Conference, WINE 2019, New York, NY, USA, December 10–12, 2019, Proceedings 15, pages 17–30. Springer, 2019

    2019

  3. [11]

    Non-quasi-linear agents in quasi-linear mechanisms

    Moshe Babaioff, Richard Cole, Jason Hartline, Nicole Immorlica, and Brendan Lucier. Non-quasi-linear agents in quasi-linear mechanisms. arXiv preprint arXiv:2012.02893, 2020

    2012 arXiv

  4. [12]

    Towards efficient auctions in an auto-bidding world

    Yuan Deng, Jieming Mao, Vahab Mirrokni, and Song Zuo. Towards efficient auctions in an auto-bidding world. In Proceedings of the Web Conference 2021, pages 3965–3973, 2021

    2021

  5. [13]

    The landscape of auto-bidding auctions: Value versus utility maximization

    Santiago R Balseiro, Yuan Deng, Jieming Mao, Vahab S Mirrokni, and Song Zuo. The landscape of auto-bidding auctions: Value versus utility maximization. In Proceedings of the 22nd ACM Conference on Economics and Computation, pages 132–133, 2021

    2021

  6. [14]

    Auction design in an auto-bidding setting: Randomization improves efficiency beyond vcg

    Aranyak Mehta. Auction design in an auto-bidding setting: Randomization improves efficiency beyond vcg. In Proceedings of the ACM Web Conference 2022, pages 173–181, 2022

    2022

  7. [15]

    Incentive compatibility in the auto-bidding world

    Yeganeh Alimohammadi, Aranyak Mehta, and Andres Perlroth. Incentive compatibility in the auto-bidding world. In Proceedings of the 24th ACM Conference on Economics and Computation, pages 63–63, 2023

    2023

  8. [16]

    Truthful auctions for automated bidding in online advertising

    Yidan Xing, Zhilin Zhang, Zhenzhe Zheng, Chuan Yu, Jian Xu, Fan Wu, and Guihai Chen. Truthful auctions for automated bidding in online advertising. In Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, pages 2915–2922, 2023

    2023

  9. [17]

    Adwords and generalized online matching

    Aranyak Mehta, Amin Saberi, Umesh Vazirani, and Vijay Vazirani. Adwords and generalized online matching. Journal of the ACM (JACM), 54(5):22–es, 2007

    2007

  10. [18]

    The adwords problem: online keyword matching with budgeted bidders under random permutations

    Nikhil R Devanur and Thomas P Hayes. The adwords problem: online keyword matching with budgeted bidders under random permutations. In Proceedings of the 10th ACM conference on Electronic commerce, pages 71–78, 2009

    2009

  11. [19]

    Learning in repeated auctions with budgets: Regret minimization and equilibrium

    Santiago R Balseiro and Yonatan Gur. Learning in repeated auctions with budgets: Regret minimization and equilibrium. Management Science, 65(9):3952–3968, 2019

    2019

  12. [20]

    Pacing equilibrium in first price auction markets

    Vincent Conitzer, Christian Kroer, Debmalya Panigrahi, Okke Schrijvers, Nicolas E Stier-Moses, Eric Sodomka, and Christopher A Wilkens. Pacing equilibrium in first price auction markets. Management Science, 68(12):8515– 8535, 2022

    2022

  13. [21]

    The complexity of pacing for second-price auctions.Mathematics of Operations Research, 2023

    Xi Chen, Christian Kroer, and Rachitesh Kumar. The complexity of pacing for second-price auctions.Mathematics of Operations Research, 2023. 13 IC Mechanisms for Risk-Averse Advertisers in the Online Advertising System

    2023

  14. [22]

    Multiplicative pacing equilibria in auction markets

    Vincent Conitzer, Christian Kroer, Eric Sodomka, and Nicolas E Stier-Moses. Multiplicative pacing equilibria in auction markets. Operations Research, 70(2):963–989, 2022

    2022

  15. [23]

    Budget pacing in repeated auctions: Regret and efficiency without convergence

    Jason Gaitonde, Yingkai Li, Bar Light, Brendan Lucier, and Aleksandrs Slivkins. Budget pacing in repeated auctions: Regret and efficiency without convergence. arXiv preprint arXiv:2205.08674, 2022

    2022

  16. [24]

    Bid shading and bidder surplus in the us treasury auction system

    Ali Hortaçsu, Jakub Kastl, and Allen Zhang. Bid shading and bidder surplus in the us treasury auction system. American Economic Review, 108(1):147–169, 2018

    2018

  17. [25]

    Adaptive bid shading optimization of first-price ad inventory

    Niklas Karlsson and Qian Sang. Adaptive bid shading optimization of first-price ad inventory. In 2021 American Control Conference (ACC), pages 4983–4990. IEEE, 2021

    2021

  18. [26]

    Mebs: Multi-task end-to-end bid shading for multi-slot display advertising

    Zhen Gong, Lvyin Niu, Yang Zhao, Miao Xu, Haoqi Zhang, Zhenzhe Zheng, Zhilin Zhang, Rongquan Bai, Chuan Yu, Jian Xu, et al. Mebs: Multi-task end-to-end bid shading for multi-slot display advertising. In Proceedings of the 32nd ACM International Conference on Information and Kn...

    2023

  19. [27]

    Meow: A space-efficient nonparametric bid shading algorithm

    Wei Zhang, Brendan Kitts, Yanjun Han, Zhengyuan Zhou, Tingyu Mao, Hao He, Shengjun Pan, Aaron Flores, San Gultekin, and Tsachy Weissman. Meow: A space-efficient nonparametric bid shading algorithm. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data ...

    2021

  20. [28]

    Model-based reinforcement learning for auto-bidding in display advertising

    Shuang Chen, Qisen Xu, Liang Zhang, Yongbo Jin, Wenhao Li, and Linjian Mo. Model-based reinforcement learning for auto-bidding in display advertising. In Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, pages 1560–1568, 2023

    2023

  21. [29]

    A personalized automated bidding framework for fairness-aware online advertising

    Haoqi Zhang, Lvyin Niu, Zhenzhe Zheng, Zhilin Zhang, Shan Gu, Fan Wu, Chuan Yu, Jian Xu, Guihai Chen, and Bo Zheng. A personalized automated bidding framework for fairness-aware online advertising. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Dat...

    2023

  22. [30]

    Auction design for roi-constrained buyers

    Negin Golrezaei, Ilan Lobel, and Renato Paes Leme. Auction design for roi-constrained buyers. In Proceedings of the Web Conference 2021, pages 3941–3952, 2021

    2021

  23. [31]

    Robust auction design in the auto-bidding world

    Santiago Balseiro, Yuan Deng, Jieming Mao, Vahab Mirrokni, and Song Zuo. Robust auction design in the auto-bidding world. Advances in Neural Information Processing Systems, 34:17777–17788, 2021

    2021

  24. [32]

    Optimal mechanisms for a value maximizer: The futility of screening targets

    Santiago Balseiro, Yuan Deng, Jieming Mao, Vahab Mirrokni, and Song Zuo. Optimal mechanisms for a value maximizer: The futility of screening targets. Available at SSRN 4351927, 2023

    2023

  25. [33]

    Auction design for bidders with ex post roi constraints

    Hongtao Lv, Xiaohui Bei, Zhenzhe Zheng, and Fan Wu. Auction design for bidders with ex post roi constraints. In International Conference on Web and Internet Economics, pages 492–508. Springer, 2023. 14 IC Mechanisms for Risk-Averse Advertisers in the Online Advertising System ...

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.