Evaluating and Pricing Advertisements in AI-Generated Responses

John L. Turner-Smith; Tonghan Wang; Yihang Zhang; Yuhan Fu; Zimeng Huang

REVIEW 4 major objections 4 minor 39 references

Evaluating and Pricing Advertisements in AI-Generated Responses

T0 review · 4 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Synthetic persona-based intent signals can replace missing click logs for AI-embedded ads, enabling a differentiable evaluator and a truthful per-click auction.

desk verdict A serious and unusually honest attempt to build a synthetic click-intent signal for LLM-native ads, with a genuinely new measurement construction; but the headline claims overstate what the evidence supports, since the 103/103 generalization test largely verifies that the evaluator learned the relevance gate baked into its own training label. read the letter →

arxiv 2607.27686 v1 pith:KNWVF2KI submitted 2026-07-30 cs.AI

John L. Turner-Smith , Zimeng Huang , Yuhan Fu , Yihang Zhang , Tonghan Wang This is my paper · ORCID

classification cs.AI

keywords click-throughintentLLM-nativeadvertisingpersona-agentsimulationdifferentiableevaluatortruthfulauctionEarthMover'sDistancebest-of-kallocationgenerativeadoptimisation

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

The paper tries to establish that click-through intent—how likely a user is to click an ad embedded inside an AI-generated answer—can be synthesised when there is no click data, no reliable human annotation, and no honest LLM judge. Its method: simulated personas with different Big Five personality profiles rate objective ad features, and those ratings are combined into a reproducible click-intent label. A compact neural evaluator trained on those labels predicts intent as a smooth, continuous score, passing behavioural tests where zero-shot LLM judges fail and generalising to invented products. The paper then uses that score as the allocation function in an auction and derives the unique payment rule under which truthful bidding is optimal. If true, this supplies the missing measurement layer that both pricing and generative ad optimisation presuppose.

What carries the argument

The central object is the persona-agent click-intent label: an LLM scores six objective ad features, and click intent is computed as copy quality times relevance/5, modulated by a bounded Big Five personality term and clipped to [0.2, 5.0], then averaged over thirty personas. That label is distilled into a shared-bottleneck neural evaluator with four prediction heads—a single shared representation feeding dimension-specific heads—trained with an ordinal loss that penalises prediction errors by their distance on the 1–5 scale. The click-intent head supplies the continuous, deterministic mapping f(q,r); taking the expected maximum of f over k candidate responses makes the allocation x(k) monot

What would settle it

Run a live A/B test in which real users are shown a sample of the evaluated ad-embedded responses and compare observed per-item click-through rates against the evaluator's predicted intent rankings; if the ordering or the relative gaps diverge systematically—especially the relevance-gate assumption that off-topic ads are never clicked—the central claim fails. A cheaper pre-registered check: test whether the evaluator's 103/103 fictional-product discrimination survives when the same products receive human click-likelihood ratings.

Watch

Extended reading notes

Core claim

Click-through intent for LLM-embedded ads can be created synthetically and then learned by a cheap, differentiable evaluator. The authors construct labels through a two-stage persona-agent simulation: an objective scorer rates six ad features, and Big Five personality traits modulate a relevance-gated base score; averaging over thirty sampled personas yields a reproducible 1–5 intent label. Training a shared-bottleneck evaluator with an ordinal score-distance loss on these labels produces a smooth expected-intent output that outperforms zero-shot frontier judges on relevance sensitivity (79% versus 60–67%), tracks dose-response degradation monotonically, generalises without error to 103 fict

Load-bearing premise

The load-bearing premise is that the synthetic label formula—copy quality times relevance divided by five, modulated by Big Five personality weights and clipped—is a faithful stand-in for real users' click-through intent; if that proxy is wrong, the evaluator and the auction are pricing the wrong quantity even though the mechanism itself is internally truthful.

Editorial extensions

If this is right

Advertisers bidding in this mechanism can be charged per click under a payment rule where truthful bidding is optimal, and the per-click price is guaranteed never to exceed the declared valuation.
Because the evaluator is differentiable and fast, the same intent signal can be used directly as a reward for training ad-generation policies, closing the generative loop that currently lacks supervision.
The evaluator's zero-error generalisation to 103 fictional products indicates the signal tracks semantic intent rather than memorised product-name associations, so it should keep working for genuinely novel ads.
The mechanism prices any measurable allocation, not just best-of-k: ironing and ε-incentive-compatible prices handle learned policies whose allocation is non-monotone, so platforms need not retrain their generators.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper's own conclusions, the relevance-gated label construction implies that an ad generator optimised against this signal will be rewarded for being on-topic and well-written above all else; a natural stress test is whether the evaluator can be gamed by superficially relevant but manipulative copy.
The paper leaves implicit that its per-click prices are not real money until the ordinal score is calibrated against live click-through rates; until that calibration exists, the auction example shows the mechanism's structure, not realised revenue.
A transferable consequence: the same persona-modulation-plus-distillation recipe could be applied to other subjective qualities that LLM judges conflate with fluency, such as trustworthiness, humour, or persuasiveness.
A live experiment could test the central premise directly: show real users a sample of the evaluated ad-embedded responses, record their clicks, and check whether observed per-item click order matches the evaluator's predicted intent rankings.

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.

Desk Editor's Note

read the letter

Worth your time. The real contribution here is the measurement layer: constructing click-intent supervision through persona-agent simulation with Big Five trait weighting, distilling it into a differentiable LoRA-adapted evaluator with an EMD loss, and validating it behaviorally. That combination is new, and the paper is transparent about its construction in a way the related work is not. The held-out fit (Q4 Pearson 0.82 on an agent-grounded target) is credible, the degradation and keyword-stuffing tests show the trained evaluator beats zero-shot judges on shape and monotonicity, and the blind pairwise human study (86% per-annotator agreement, 92% majority, rising with confidence) is decent initial evidence that the evaluator's orderings are sensible. The mechanism section is correct but standard: the payment identity is Myerson, and the best-of-k example is a clean demonstration, not a new theoretical result. The authors mostly say this themselves.

The soft spot is the one the reader flagged, and it is load-bearing. The Q4 label is defined by a hand-set formula over LLM feature scores with a relevance gate, so every downstream result inherits whatever that formula measures. The fictional-product test varies relevance while holding the ad fixed, and Eq. (3) makes the label monotonically increasing in relevance by construction. Getting 103/103 there means the evaluator learned the gate, not that it internalised click intent beyond relevance. The cross-category swap test has the same confound. The paper's own Section 6 concedes the labels are an ordinal surrogate and have not been validated against an independent human standard at scale, and that the scale is not calibrated to real CTR. None of this sinks the framework as a reproducible, useful ordinal signal for generation selection and mechanism illustration, but it does mean the abstract's 'predicts click-through intent' and 'internalised a meaningful intent signal' are stronger than what the experiments establish. The perturbation N is small (80 items, 15 swap pairs), which is a minor additional concern given how much weight rests on those tests.

The authors deserve credit for flagging the validation gap themselves, and the mechanism design is internally sound conditional on the evaluator. This paper should get a serious referee, not a desk reject. A referee should push for either independent human validation at scale or live A/B data, and should require the claims in the abstract to be softened to match the evidence. If it ships code and data, I would cite it as the cold-start CTI evaluation framework to beat. Bring it to reading group; it will spark a good argument.

Referee Report

4 major / 4 minor

Summary. The paper addresses the absence of click-through intent (CTI) supervision for advertisements embedded in LLM-generated responses. It constructs synthetic labels via a persona-agent simulation in which an LLM scores six ad features and a hand-specified formula combines them with Big Five personality modulation and a relevance gate (Eqs. 1–3). These labels, together with NaiAD human labels for the other three quality dimensions, supervise a shared-bottleneck evaluator built on a frozen Qwen3-4B backbone with LoRA and an EMD loss. The authors report held-out Q4 Pearson correlation 0.82, a battery of six behavioral perturbation tests, 103/103 fictional-product relevance discrimination, and 86% mean per-annotator pairwise agreement with human judges in Best-of-N selection. They then use the evaluator output as the monotone allocation primitive in a Myerson-style truthful auction, deriving the payment identity, a best-of-k worked example, and extensions to ironing and ε-incentive compatibility. Section 6 is explicit that the agent labels are an uncalibrated ordinal surrogate not yet validated against a large independent human standard.

Significance. If the synthetic CTI label were a valid proxy for real user engagement, the paper would supply a genuinely useful primitive: a deterministic, differentiable, and cheap CTI estimator usable for both mechanism design and as a reward for generative ad optimization. The mechanism-design layer is standard but competently applied, and the persona-agent labelling framework is a reproducible methodological proposal. The paper is also unusually candid about its limitations. However, the current evidence does not establish that the evaluator measures click-through intent rather than the relevance-gated formula used to train it. The headline fictional-product and swap results are largely re-encodings of Eq. (3), and the human validation is small in scale. The contribution is therefore conditional: as a proposal for constructing synthetic intent supervision it is valuable; as a validated CTI evaluator it overclaims. I would like to see the claims scaled back or a substantially stronger external-validation study.

major comments (4)

[§4.3, test 6; Eq. (3)] The 103/103 fictional-product result does not support the claim that the evaluator has internalised a meaningful intent signal beyond relevance. In this test the same advertisement is placed in a contextually relevant and an irrelevant context; because Eq. (3) defines the label as c = clip(copy·(relevance/5)·(1+λ m_p), 0.2, 5.0), for fixed copy and personality the label is monotonically increasing in relevance. An evaluator that merely learns the relevance-gate achieves 103/103. The cross-category swap test (test 1) has the same confound: replacing the ad with an off-category product changes only topical relevance. The sentence 'the discrimination is necessarily semantic' should read 'necessarily semantic-relevance.' To support the stronger intent claim, add perturbations that vary copy quality or personality while holding relevance constant, or that require combining relevance with the
[§3.1 and §6] The entire downstream chain inherits the validity of the synthetic label formula, but the formula is a hand-assembled construct with free parameters (λ=0.5, trait–feature weights s_{t,f}, K=3, K_personas=30, clip bounds, affine transform (f−1)/4). Section 6 concedes that the agent-grounded labels have not been validated against a large independent human standard and are 'an ordinal surrogate for engagement, not a calibrated probability.' The Section 4.4 human study is small (100 pairs, five annotators, Fleiss κ=0.64) and measures judged click likelihood rather than real engagement. Consequently the abstract's claim of a validated click-intent signal is not supported. The authors should either provide independent validation linking the labels or evaluator to real or externally elicited click preferences, or explicitly restrict all claims to a 'synthetic intent proxy.'
[§4.3, tests 1–5] The sign-certain perturbation battery is internal consistency testing, not external validation. Because the training labels are constructed to be monotonically sensitive to relevance and copy quality, an evaluator that has learned the training formula will by construction lower scores on off-category swaps, filler corruption, and keyword stuffing. The comparison with zero-shot judges is informative about relative sensitivity, but it does not validate that the labels track human click intent. The paper's phrase 'intrinsically validated' is appropriate; the abstract's stronger claim that the evaluator has 'internalised a meaningful intent signal' goes beyond what these tests can show.
[§5, Eq. (6)] The auction results are mathematically correct for any monotone allocation function x(v), but their economic meaning depends on f being a calibrated or at least valid measure of click probability. Since f is an uncalibrated ordinal score (as acknowledged in Section 6), the per-click price π(v) is a price per synthetic intent unit, not per expected real click. The worked example notes this, but the abstract and Section 5's framing ('truthful pricing') should make equally explicit that truthfulness is with respect to the learned synthetic signal, pending calibration against live engagement data.

minor comments (4)

[§4.4] Report confidence intervals or significance tests for the 86%/92% agreement figures; with 100 pairs and five annotators the estimates are noisy. Also clarify whether the 86% is the mean of per-annotator agreement or a pooled agreement.
[§3.1] A sensitivity analysis for the free parameters λ and s_{t,f} would strengthen the paper. The authors state the weights can be re-tuned post hoc, but do not show how the evaluator's behavior changes under plausible re-tunings.
[Abstract and §4.3] The phrase 'generalises without error to 103 fictional products' should be qualified as 'distinguishes relevant from irrelevant placements of the same ad in 103/103 fictional product categories.' As written, it invites an overly broad reading.
[Table 1] The explanation for Q1's low Spearman ('two-thirds of items fall within a 0.2-wide band') should state which quantity falls in that band (presumably the label values).

Circularity Check

1 steps flagged · score 6.0 of 10

The 103/103 fictional-product and cross-category swap results are entailed by the relevance gate in the paper's own label formula; the 'internalised intent signal' claim is therefore partly a restatement of the training target, with only partial external support from the small pairwise human study.

self definitional [Section 3.1, Eqs. (2)-(3); Section 4.3, Test 6 (Fictional-product generalisation)]
"basep = copy·(relevance/5), cp = clip(basep ·(1 + λ mp), 0.2, 5.0) ... The gate ensures that an irrelevant advertisement cannot earn high click-through intent regardless of copy quality or personality. ... for each of 103 fabricated product categories ... one contextually relevant and one irrelevant placement of the same advertisement were generated. The evaluator assigns the relevant placement the higher click-intent score in all 103 cases."

In Test 6 the advertisement is identical in the relevant and irrelevant placements, so copy, personality weights, and all non-relevance features are fixed. The only input that varies in the training label of Eq. (3) is relevance, and c_p is monotonically increasing in relevance by construction. Any model that fits its supervision must therefore rank the relevant placement higher; the 103/103 score and the 79% swap pass rate are checks that the evaluator learned the relevance gate built into its own labels, not independent evidence of click-through intent. The paper's Section 6 concession that the scores are 'an ordinal surrogate for engagement, not a calibrated probability' underlines that this test validates the synthetic target, not a real-world intent signal.

full rationale

The pricing derivation is not circular: x(k) = E[max f(r_j)] is monotone by construction as an expected maximum, and the payment identity p(v) = v x(v) - ∫ x(t)dt is the standard TIE characterization; this is a theorem, not an input. The circularity is concentrated in the validation claims about click-intent. With the ad fixed, the agent label c_p depends on relevance only through the product copy·(relevance/5), so the training target itself orders the relevant placement above the irrelevant one. The evaluator is trained on these labels with an EMD loss, so it can pass the fictional-product and cross-category swap tests by learning its own target's relevance gate. The paper even concedes in Section 6 that the evaluator 'tracks semantic relevance and text quality first' with psychological features as a bounded modulation. The only external anchor is the blind pairwise human study, which is small (100 pairs, five annotators) and is described by the authors as 'initial evidence' rather than scale validation. The NaiAD self-citation about uncalibratable human CTI is load-bearing for the choice of synthetic labels but is an empirical premise rather than a definitional reduction; it contributes to the weakness but does not alone create the circularity. Overall, some 'predictions' reduce by construction, but the mechanism-design result and the human pairwise anchor give the paper partial independent content, so the score is 6, not higher.

Assumptions & free parameters 5 free parameters · 7 assumptions · 2 invented entities

The central construction rests on hand-set formulas and LLM-generated feature scores; the only independent empirical anchor is a small human pairwise study. The pricing layer adds standard mechanism-design axioms.

free parameters (5)

λ (personality modulation strength) = 0.5
Hand-set in Eq. (3); controls how much personality can move the click-intent label. No fitting procedure given.
signed trait–feature weights s_{t,f} = sign-fixed, magnitude unspecified (appears to be ±1)
Used in Eq. (1); directions from literature, magnitudes not estimated. Re-tuning is advertised as possible post hoc (§6).
K=3 and K_personas=30 = 3, 30
Averaging counts chosen by hand; no sensitivity analysis.
clip bounds 0.2 and 5.0 = 0.2, 5.0
Arbitrary floor/ceiling in Eq. (3) that shapes the label distribution.
affine transform (f−1)/4
Maps 1–5 evaluator output to [0,1] for pricing; acknowledged in §6 as not calibrated to absolute click frequencies.

assumptions (7)

domain assumption LLM feature scoring in Stage 1 is a reliable, consistent measure of objective ad quality features.
Section 3.1 Stage 1; the entire label rests on these six scores from a single LLM (Qwen family), averaged K=3.
ad hoc to paper Relevance gate: an off-topic ad cannot receive high click-through intent regardless of copy quality or personality.
Eq. (2): base = copy·(relevance/5). This is a design prior, not an empirical regularity; it also shapes the fictional-product generalization test.
domain assumption Big Five trait–feature pairings and their signed effects apply to LLM-native advertising.
Section 3.1 Stage 2, citing psychology literature; authors note uneven support and meta-analytic caution (§6).
domain assumption Normative Big Five population distributions from cited agent simulation work yield realistic personas.
Section 3.1 'sample K personas = 30 from normative human population distributions' citing Zhou et al. 2025.
domain assumption Sign-certain perturbation directions are known a priori (foreign ad lower, keyword stuffing lower, degradation lower).
Section 4.3 relies on these directional priors to validate the evaluator; they align with the label formula itself.
domain assumption NaiAD VC-PPI human labels for Q1–Q3 are valid supervision.
Section 3.2 uses them as targets; relies entirely on cited benchmark (same research group).
standard math Standard mechanism-design characterization: a single-parameter mechanism is TIE iff x(v) is monotone, with payment identity p(v)=vx(v)−∫x.
Theorem A.2 in supplementary, Myerson-style; independent of empirical claims.

invented entities (2)

Persona-agent click-intent labels
purpose: Synthetic supervision for Q4 in place of unavailable click logs and uncalibratable human labels.
Defined by Eqs. (1)–(3) from LLM feature scores and hand-set personality weights; no real-world click data used. Human pairwise study gives weak external evidence but no ground truth.
Click-through intent (CTI) as a continuous latent score on 1–5
purpose: Differentiable signal for pricing and generation.
Acknowledged in §6 as an ordinal surrogate, not a calibrated CTR probability; no live-data validation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Evaluating and Pricing Advertisements in AI-Generated Responses." pith.science (2026). https://pith.science/paper/KNWVF2KI

@misc{pith2026260727686,
  author       = {Pith},
  title        = {Pith review of: Evaluating and Pricing Advertisements in AI-Generated Responses},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KNWVF2KI}},
  note         = {Machine review of arXiv:2607.27686}
}

read the original abstract

As search increasingly shifts toward LLM-driven answer engines, advertising is becoming embedded within the generated response itself and should therefore be evaluated for both user utility and commercial value. The key challenge is click-through intent: behavioural logs are unavailable, human annotation resists calibration, and frontier LLM judges conflate intent with linguistic fluency. These gaps compound, as principled pricing presupposes a continuous intent signal, while generating such a signal presupposes supervision that is currently unavailable. We construct the missing supervision through a psychologically grounded agent simulation framework, and distil it into a parameter-efficient evaluator that predicts click-through intent, together with the three companion dimensions of ad quality, as smooth, differentiable estimates. Validated through sign-certain behavioural perturbations, the evaluator surpasses frontier zero-shot judges on relevance sensitivity (79% versus 60-67%), tracks graded content degradation, generalises without error to 103 fictional products, and agrees with human preference in 86% of pairwise judgements across five annotators, with agreement rising in the evaluator's confidence. Upon its estimates we build the pricing layer directly, deriving the unique payment rule under which truthful bidding is optimal, demonstrating it on a best-of-k allocation, and extending the mechanism to non-monotone allocations. The same differentiable signal stands ready as a training objective for ad generation.

Figures

Figures reproduced from arXiv: 2607.27686 by the authors.

**Figure 1.** Overview of the framework. From the absence of reliable click-intent supervision (left), a persona-agent simulation [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** Behavioural comparison with zero-shot judges. (a) Mean predicted click-intent as an increasing fraction of the ad [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 1 linked inside Pith

[1]

Zhang, Yihang and Huang, Zimeng and Zhai, Ren and Kang, Yipeng and Wang, Tonghan , journal=
[2]

Influence of personality traits on generation

Saha, Partha and Sengupta, Angan and Gupta, Priya , journal=. Influence of personality traits on generation. 2024 , publisher=

2024
[3]

Simulating Prosocial Behavior and Social Contagion in

Zhou, Yujia and Wang, Hexi and Ai, Qingyao and Wu, Zhen and Liu, Yiqun , journal=. Simulating Prosocial Behavior and Social Contagion in
[4]

Scaling Personality Control in

Cho, Gunhee and Cheong, Yun-Gyung , journal=. Scaling Personality Control in
[5]

Online advertisements with

Feizi, Soheil and Hajiaghayi, MohammadTaghi and Rezaei, Keivan and Shin, Suho , journal=. Online advertisements with
[6]

Proceedings of the ACM Web Conference 2024 , pages=

Mechanism design for large language models , author=. Proceedings of the ACM Web Conference 2024 , pages=

2024
[7]

Truthful aggregation of

Soumalias, Ermis and Curry, Michael and Seuken, Sven , booktitle=. Truthful aggregation of
[8]

Ad auctions for

Hajiaghayi, MohammadTaghi and Lahaie, S. Ad auctions for. Advances in Neural Information Processing Systems , volume=

Show all 39 references

[9]

Auctions with

Dubey, Avinava and Feng, Zhe and Kidambi, Rahul and Mehta, Aranyak and Wang, Di , booktitle=. Auctions with
[10]

Hu, Silan and Zhang, Shiqi and Shi, Yimin and Xiao, Xiaokui , journal=
[11]

Ad Insertion in

Xu, Shengwei and Chen, Zhaohua and Deng, Xiaotie and Huang, Zhiyi and Schoenebeck, Grant , journal=. Ad Insertion in
[12]

Zhao, Chujie and Hu, Qun and Song, Shiping and Chen, Dagui and Zhu, Han and Xu, Jian and Zheng, Bo , journal=
[13]

Ads that Talk Back: Implications and Perceptions of Injecting Personalized Advertising into

Tang, Brian Jay and Sun, Kaiwen and Curran, Noah T and Schaub, Florian and Shin, Kang G , journal=. Ads that Talk Back: Implications and Perceptions of Injecting Personalized Advertising into
[14]

Proceedings of the 1st Workshop on Deep Learning for Recommender Systems , pages=

Wide & deep learning for recommender systems , author=. Proceedings of the 1st Workshop on Deep Learning for Recommender Systems , pages=
[15]

Guo, Huifeng and Tang, Ruiming and Ye, Yunming and Li, Zhenguo and He, Xiuqiang , journal=
[16]

Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

Deep interest network for click-through rate prediction , author=. Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=
[17]

2010 IEEE International Conference on Data Mining , pages=

Factorization machines , author=. 2010 IEEE International Conference on Data Mining , pages=. 2010 , organization=

2010
[18]

Ma, Luyi and Zhang, Wanjia Sherry and Fan, Zezhong and Thakur, Shubham and Zhao, Kai and Yao, Kehui and Agarwal, Ayush and Iyer, Rahul and Cho, Jason and Xu, Jianpeng and others , journal=
[19]

Proceedings of the 13th ACM Conference on Electronic Commerce , pages=

Social influence in social advertising: Evidence from field experiments , author=. Proceedings of the 13th ACM Conference on Electronic Commerce , pages=
[20]

Journal of Advertising , volume=

Explaining the impact of scarcity appeals in advertising: The mediating role of perceptions of susceptibility , author=. Journal of Advertising , volume=
[21]

Political Analysis , volume=

Out of one, many: Using language models to simulate human samples , author=. Political Analysis , volume=
[22]

Psychology & Marketing , volume=

Using large language models to generate silicon samples in consumer and marketing research: Challenges, opportunities, and guidelines , author=. Psychology & Marketing , volume=
[23]

Human Psychometric Questionnaires Mischaracterize

Song, Woojung and Choi, Dongmin and Park, Yoonah and Han, Jongwook and Jo, Yohan , journal=. Human Psychometric Questionnaires Mischaracterize
[24]

and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , booktitle=

Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , booktitle=
[25]

International Conference on Learning Representations , year=

Flow matching for generative modeling , author=. International Conference on Learning Representations , year=
[26]

arXiv preprint arXiv:1611.05916 , year=

Squared earth mover's distance-based loss for training deep neural networks , author=. arXiv preprint arXiv:1611.05916 , year=

arXiv
[27]

Beyond accuracy: Behavioral testing of

Ribeiro, Marco Tulio and Wu, Tongshuang and Guestrin, Carlos and Singh, Sameer , booktitle=. Beyond accuracy: Behavioral testing of
[28]

IEEE Transactions on Software Engineering , volume=

A survey on metamorphic testing , author=. IEEE Transactions on Software Engineering , volume=
[29]

ACM Computing Surveys , volume=

Metamorphic testing: A review of challenges and opportunities , author=. ACM Computing Surveys , volume=
[30]

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , pages=

The ``problem'' of human label variation: On ground truth in data, modeling and evaluation , author=. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , pages=

2022
[31]

Transactions of the Association for Computational Linguistics , volume=

Dealing with disagreements: Looking beyond the majority vote in subjective annotations , author=. Transactions of the Association for Computational Linguistics , volume=
[32]

Proceedings of the AAAI Conference on Artificial Intelligence , volume=

Toward a perspectivist turn in ground truthing for predictive computing , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=
[33]

IEEE Transactions on Neural Networks and Learning Systems , volume=

Learning from noisy labels with deep neural networks: A survey , author=. IEEE Transactions on Neural Networks and Learning Systems , volume=
[34]

Psychological Science , volume=

Personalized persuasion: Tailoring persuasive appeals to recipients' personality traits , author=. Psychological Science , volume=
[35]

Proceedings of the

Psychological targeting as an effective approach to digital mass persuasion , author=. Proceedings of the
[36]

How does personality affect perception of advertising messages?

Grochowska, Alicja and M. How does personality affect perception of advertising messages?. International Journal of Advertising , volume=
[37]

Perla, Raphael and Maran, Thomas and Bagci, Beg. The (. Psychology & Marketing , volume=
[38]

Mathematics of Operations Research , volume=

Optimal auction design , author=. Mathematics of Operations Research , volume=
[39]

Zhang, Zhaowei and Fu, Yuhan and Zhang, Yihang and Liu, Xiaohan and Zhang, Ceyao and Zhang, Xiaoyuan and Kang, Yipeng and Wang, Tonghan and Yang, Yaodong , journal=

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Zhang, Yihang and Huang, Zimeng and Zhai, Ren and Kang, Yipeng and Wang, Tonghan , journal=

[2] [2]

Influence of personality traits on generation

Saha, Partha and Sengupta, Angan and Gupta, Priya , journal=. Influence of personality traits on generation. 2024 , publisher=

2024

[3] [3]

Simulating Prosocial Behavior and Social Contagion in

Zhou, Yujia and Wang, Hexi and Ai, Qingyao and Wu, Zhen and Liu, Yiqun , journal=. Simulating Prosocial Behavior and Social Contagion in

[4] [4]

Scaling Personality Control in

Cho, Gunhee and Cheong, Yun-Gyung , journal=. Scaling Personality Control in

[5] [5]

Online advertisements with

Feizi, Soheil and Hajiaghayi, MohammadTaghi and Rezaei, Keivan and Shin, Suho , journal=. Online advertisements with

[6] [6]

Proceedings of the ACM Web Conference 2024 , pages=

Mechanism design for large language models , author=. Proceedings of the ACM Web Conference 2024 , pages=

2024

[7] [7]

Truthful aggregation of

Soumalias, Ermis and Curry, Michael and Seuken, Sven , booktitle=. Truthful aggregation of

[8] [8]

Ad auctions for

Hajiaghayi, MohammadTaghi and Lahaie, S. Ad auctions for. Advances in Neural Information Processing Systems , volume=