NLGR: Utilizing Neighbor Lists for Generative Rerank in Personalized Recommendation Systems

Chi Wang; Qi Tang; Senjie Kou; Shuli Wang; Wenshuai Chen; Xingxing Wang; Xiong Xiao; Xue Wei; Yinhua Zhu

REVIEW 4 major objections 6 minor 40 references

NLGR: Utilizing Neighbor Lists for Generative Rerank in Personalized Recommendation Systems

T0 review · 4 major / 6 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a generative reranker trained with neighbor-list relative rewards and a sampling-based non-autoregressive decoder outperforms existing reranking methods on both a public and an industrial dataset, and that it lifts…

desk verdict A useful industrial neighbor-list training trick, but the headline offline HR gains are computed by the same evaluator that trained the generator, so only the online A/B test is independent evidence. read the letter →

arxiv 2502.06097 v2 pith:2LYANST3 submitted 2025-02-10 cs.IR cs.AI

Shuli Wang , Xue Wei , Senjie Kou , Chi Wang , Wenshuai Chen , Qi Tang , Yinhua Zhu , Xiong Xiao

show 1 more author

Xingxing Wang

This is my paper · ORCID

classification cs.IRcs.AI

keywords generativererankingneighborlistsevaluator-generatorparadigmnon-autoregressivegenerationGumbel-softmaxlistwiserecommendationCTRpredictiononlineA/Btest

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

This paper proposes a reranking method, NLGR, for recommender systems that reorders an initial list of candidates to maximize list-level user engagement. It claims that training the generator on 'neighbor lists'—lists that differ from the current one by a single item—resolves a goal inconsistency between the evaluator and the generator, so the generator learns to move toward better regions of the combinatorial space rather than imitating historical exposure. It also claims that a sampling-based non-autoregressive decoder, which picks an item position and then a replacement item, explores that space more flexibly than one-by-one autoregressive generation. The authors report consistent offline gains over six baselines on the Taobao Ad and Meituan datasets, and an online A/B test on Meituan with CTR up 3.25% and GMV up 3.07%.

What carries the argument

The central object is the neighbor list: a list that differs from the current candidate list by exactly one item. The relative reward $r_j - r_o$ for each position $j$, computed by the evaluator NLGR-E, converts the evaluator's listwise scores into a differentiable training signal that tells the generator which single-item move improves utility. The generator's sampling-based non-autoregressive decoder consists of the Position Decision Unit (PDU), which chooses the position to replace via Gumbel-softmax sampling, and the Candidate Retrieval Unit (CRU), which chooses the replacement item, allowing the model to jump to any neighbor list in one step.

What would settle it

Serve the lists generated by NLGR-G to a small fraction of online traffic and compare NLGR-E's predicted reward for each served list with the actual observed click and conversion behavior; if the rank correlation is weak for these generated lists, the relative-reward training signal is not a faithful proxy for user utility.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the goal inconsistency between evaluator and generator in generative reranking can be resolved by training the generator on neighbor lists: for each candidate list, the evaluator scores the list and each single-item replacement, and the generator's objective is to maximize the sum of relative rewards. The paper also claims that generating a replacement by first sampling the position to change (Position Decision Unit) and then sampling the item to insert (Candidate Retrieval Unit), both with Gumbel-softmax, yields a non-autoregressive search that jumps flexibly between neighbor lists and outperforms one-by-one sequence generation. It reports consistent offline gains over all baselines on Taobao Ad and Meituan, and an online A/B test on Meituan showing CTR +3.25% and GMV +3.07%.

Load-bearing premise

The load-bearing premise is that NLGR-E's estimated list rewards are faithful proxies for real user utility when applied to lists that were never shown to users; if the evaluator is biased on these counterfactual lists, the generator may simply overfit to the evaluator, and the offline and possibly online gains would not reflect genuine user satisfaction.

Editorial extensions

If this is right

Only the generator is served online; the evaluator is used only to train it, so deploying NLGR adds no new online evaluation cost beyond the generator's own inference.
The neighbor-list construction requires $m$ evaluator calls per training list, which is tractable for realistic list sizes (e.g., $m=4$ on Meituan, $m=5$ on Taobao Ad).
The method's gains are consistent across two datasets of very different scale and feature richness, suggesting the training signal transfers to diverse recommender settings.
The online A/B result implies that the offline generator improvements translate into commercially meaningful user engagement and revenue at industrial scale.

Reading between the lines

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

The same neighbor-list relative-reward training could apply to other combinatorial list-optimization tasks, such as ad slate optimization or search result diversification, wherever a differentiable evaluator exists.
A testable extension is to replace the hand-designed reward combination (the business parameters $k_1$ and $k_2$) with a learned reward, which could remove the need for per-business tuning.
Because the offline HR metric is computed by ranking generated lists with the same NLGR-E, the offline gains partly measure generator-evaluator consistency; the online A/B test is the cleaner evidence that real user utility improved.

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

Short version: the neighbor-list relative-reward training and the PDU/CRU sampling decoder are genuinely new combinations, and the Meituan deployment gives the paper real industrial weight. The headline offline HR numbers, though, should not be read as evidence about user utility. Section 5.1.3 says it plainly: HR measures consistency between evaluator and generator, and "the generator cannot be fully and fairly evaluated." Since NLGR-G is trained against NLGR-E's rewards (Eqs. 15-16) and then graded by NLGR-E in Table 3, the big HR lifts (0.4091 vs 0.2160 on Taobao, 0.8369 vs 0.7573 on Meituan) are mostly self-consistency. The evaluator itself sees only exposed lists in training, while the neighbor lists it scores are counterfactual; there's no validation of NLGR-E on those lists. That's the load-bearing gap.

What the paper does well: it identifies a real goal-inconsistency problem, and the relative-reward trick is a sensible way to give the generator a directional signal in combinatorial space rather than a single scalar. The sampling-based non-autoregressive decoder is a reasonable efficiency-motivated alternative to sequential decoding, and the ablations show both pieces matter. The online A/B test is independent and positive (CTR +3.25%, GMV +3.07%), though it compares against a PRM variant rather than GRN/DCDR, and there are no confidence intervals. The authors also deserve credit for stating the HR limitation explicitly instead of burying it.

Minor soft spots: NAR4Rec is cited in related work but missing from the baseline table; temperature tau and the stop condition are not specified; no code. None of these change the central method, but they make replication harder.

Who this is for: practitioners working on listwise reranking in production, and anyone studying evaluator-generator training loops. The method is worth taking seriously, but the paper needs an independent offline evaluation—a separate reward model, or some counterfactual ground truth—before the strong claims in Section 5.2.1 are supported. If this crossed my desk, I'd send it to review; the idea is useful and the deployment is real, but I'd ask for that extra evaluation or a tone-down of the claims.

Referee Report

4 major / 6 minor

Summary. The paper proposes NLGR, a generative reranking framework in the evaluator-generator paradigm. It introduces two main ideas: (i) a neighbor-list based training procedure in which the generator NLGR-G is trained using relative rewards computed by an evaluator NLGR-E on the current list and its single-item-replacement neighbor lists, and (ii) a sampling-based non-autoregressive generation method comprising a Position Decision Unit (PDU) and a Candidate Retrieval Unit (CRU) that flexibly jumps to neighbor lists. Experiments are conducted on the Taobao Ad public dataset and a large Meituan industrial dataset, with offline metrics (AUC, LogLoss, NDCG, HR) and an online A/B test on Meituan. The paper claims significant and consistent improvement over six baselines in all offline metrics on both datasets, and online CTR +3.25% and GMV +3.07% versus a PRM-based baseline, with deployment on the Meituan platform.

Significance. The neighbor-list relative reward mechanism targets a real limitation of evaluator-generator reranking paradigms, and the non-autoregressive sampling decoder is a plausible efficiency-oriented alternative to sequential decoding. If the reported results are valid, the method would be a practically useful contribution, especially given the successful deployment at Meituan. However, the primary offline metric for the generator (HR) is computed by the same evaluator that supplies the training rewards, creating a circular evaluation. The online A/B test is independent but lacks statistical detail and does not compare against the generative baselines. These issues currently temper the strength of the claims, though the core ideas are worth publishing if the validation is strengthened.

major comments (4)

[Section 5.1.3, Table 3] The HR metric used to evaluate NLGR-G is computed by ranking the generated lists with NLGR-E, the same evaluator that provides the training rewards in Eqs. (15) and (16). NLGR-G is explicitly optimized to maximize NLGR-E's estimated reward, so HR measures evaluator-generator consistency rather than user utility. The paper itself acknowledges this in Section 5.1.3 ('the generator cannot be fully and fairly evaluated'). This circularity directly undermines the central claim in Section 5.2.1 that NLGR 'significantly and consistently outperforms the state-of-the-art approaches in all metrics on both datasets,' because the offline generator gains (e.g., HR@10% of 0.8369 on Meituan) are in large part a measure of fit to the training evaluator. Please provide an independent evaluation of the generator, such as online A/B tests against GRN and DCDR with proper statistical reporting, or a counterfactual validation of NLGR-E on held-out lists with ground-truth user feedback.
[Section 4.3.1, Section 4.3.2] NLGR-E is trained only on exposed lists (real user logs), yet it is applied to score neighbor lists in the counterfactual space that were never shown to users. The paper does not validate NLGR-E's reliability under this distribution shift. Since both the training rewards and the offline HR metric depend entirely on this extrapolation, the central derivation rests on an untested assumption. Please add an analysis of NLGR-E's calibration or accuracy on counterfactual lists, for example using logged bandit data with propensity scores or a small online validation where counterfactual lists are exposed to a control group.
[Section 5.2.1, Table 2] Table 2 reports AUC, LogLoss, and NDCG for all models, but Section 5.1.3 states that these metrics evaluate NLGR-E, not the generator. It is unclear whether the AUC/NDCG values for the baselines are computed using the same NLGR-E as a scoring function, using each model's own internal scores, or using ground-truth labels on the reranked lists. A fair comparison requires that the same evaluation protocol be applied to all methods. Please clarify the exact computation and ensure the comparison is apples-to-apples; if AUC/NDCG are computed on NLGR-E's predictions, they mainly measure the quality of the evaluator, which is a different claim than the reranking performance of the full system.
[Section 5.4, Table 6] The online A/B test is the only independent evidence of real-world improvement, but it is reported without confidence intervals, p-values, or details on the traffic split and number of users. The reported gains (CTR +3.25%, GMV +3.07%) are compared only against a 'variant of PRM' and not against the generative baselines GRN or DCDR. Given that the offline metric is circular, the online experiment carries the main burden of proof for the paper's practical claims. Please include full statistical details and, if feasible, an online comparison with a stronger generative baseline.

minor comments (6)

[Throughout] There are multiple typos: 'Metuan' should be 'Meituan' in Tables 2, 4, 5 and in Section 5.2.1; Eq. (14) uses 'NLGC-E' instead of 'NLGR-E'; Section 5.1.2 has a missing space in 'self-attention. mechanism'.
[Section 4.2.1, Eq. (7)] The symbol $n$ is used both for the Gumbel noise and for the size of the candidate set in Section 4.2.2; please use distinct notation to avoid confusion.
[Section 4.2.2, Eq. (11)] The sentence 'the newly inserted item is $c = \argmax(r^c_k)$' reuses $c$ which already denotes the candidate representation; use $k^*$ or a different variable for the selected index.
[Section 5.1.4] The Gumbel-softmax temperature $\tau$ is introduced in Eq. (7) but its value is not reported in the implementation details; please add it.
[Section 5.4] The stop condition in Section 4.2.2 ('or the values of $r^p_j$ and $r^c_k$ are too low') is vague; specify the threshold or mechanism used in practice.
[References] References [1] and [2] are duplicates of the same DLCM paper, and references [36] and [37] are duplicates of the same GOMIA paper; please consolidate.

Circularity Check

2 steps flagged · score 6.0 of 10

Offline HR gains are a closed loop: NLGR-G is trained to maximize NLGR-E's reward and then graded by the same NLGR-E; the paper itself calls HR a consistency measure, so the headline offline superiority claim is only partly independent.

fitted input called prediction [Section 4.3.2 (Eqs. 15-16) and Section 5.1.3]
"It is worth noting that only one list produced by reranking algorithms can be presented to the user. As a result, the generator cannot be fully and fairly evaluated. A practical workaround is to employ the evaluator to assess the performance of the generator. For each data, we evaluate all candidate lists using NLGR-E. HR@10% is 1 only when the rerank list produced by NLGR-G is ranked within the top 10% as sorted by NLGR-E."

NLGR-G's training objective (Eq. 16) is L^G_1 = -R = - sum r_j, where r_j = r_j - r_o are relative rewards produced by NLGR-E (Eqs. 13-15). The offline HR@10% is then computed by ranking candidate lists with the same NLGR-E. Thus the generator is fitted to NLGR-E's preferences and the metric reports how well the fit succeeded. High HR is a measure of self-agreement, not an externally grounded list-utility measure. The paper itself concedes 'the generator cannot be fully and fairly evaluated' but Section 5.2.1 uses HR gains ('0.8369/0.4091 absolute HR@10%') as primary evidence of 'extreme improvements via counterfactual evaluation' and of outperforming SOTA in all metrics.
self definitional [Section 5.1.3]
"AUC measures the model’s ability to evaluate an ordered list, while HR measures the consistency between the evaluator and generator. A decrease in any indicator will reduce the recommendation effect."

The paper defines HR as 'the consistency between the evaluator and generator' and in the same section uses that HR as the offline measure of NLGR-G's performance. Since NLGR-G's loss is explicitly the negative of NLGR-E's estimated reward, the consistency is partly secured by construction: optimizing against an evaluator and then ranking by that evaluator cannot falsify the generator's utility independently. The statement that a decrease in HR 'will reduce the recommendation effect' assumes, rather than shows, that evaluator-generator agreement equals recommendation quality on never-exposed lists.

full rationale

The paper is not wholly circular: NLGR-E is trained on real exposed-list click/conversion labels (Eq. 12), so the evaluator has independent predictive content; Table 2's AUC/LogLoss/NDCG evaluate that aspect; and the online A/B test (CTR +3.25%, GMV +3.07%) is an external check against a PRM variant. However, the offline 'state-of-the-art in all metrics' claim rests substantially on HR@10%, which is a closed loop: the generator's training rewards and the HR grading share the same NLGR-E. The paper's own limitation statement ('generator cannot be fully and fairly evaluated') confirms that HR is not an independent performance measure. Because the central offline superiority claim is partially by construction, the score is 6 rather than 0-2. No load-bearing self-citation was found, and the online A/B test prevents a higher score.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central result rests on the counterfactual validity of the evaluator, the sufficiency of single-item swap neighborhoods, and standard Gumbel-softmax training. No new conceptual entities are introduced, but several free parameters (k1, k2, alpha, beta, tau) affect the training signal and are tuned or set ad hoc, which increases the burden on the empirical evidence.

free parameters (4)

k1, k2 (reward conversion weights) = not reported, business-dependent
In Eq. 14, the evaluator's pCTR and pCVR are combined into a business reward with weights k1 and k2. They are set by business considerations (click bid and conversion price) and directly determine the reward signal used to train the generator.
alpha (loss weight) = 0.2
In Eq. 18, alpha balances the counterfactual list loss and the PDU cross-entropy loss. The value 0.2 is chosen after hyperparameter search reported in Table 5.
beta (neighbor sampling ratio) = 1
Controls the number of neighbor lists sampled per position during training (Section 5.3). Swept in Table 5; default 1 is used in the main results.
Gumbel-softmax temperature tau = not reported
Appears in Eqs. 7 and 11 for sampling replaced positions and replacement items. The paper never states the value, which affects exploration and gradient estimation.

assumptions (3)

domain assumption The evaluator NLGR-E, trained on logged exposure data, provides valid rewards for unexposed neighbor lists (counterfactual validity).
Section 4.3.2 uses NLGR-E to score neighbor lists that were never shown to users. This extrapolates beyond the training distribution, and the paper does not validate these counterfactual scores against ground truth.
domain assumption A single-item replacement neighborhood (distance 1, Section 3) is sufficient for the generator to navigate to a near-optimal list.
The PDU/CRU decoder moves one item at a time and training rewards are one-step neighbor deltas. No convergence guarantee is provided for this local search process.
standard math Gumbel-softmax gives usable gradients for discrete sampling.
Used in Eqs. 7 and 11 following prior work [15, 17, 23]; the paper relies on the standard reparameterization without discussing bias or variance.

how reviews work

0 comments

Cite this review

Pith. "Pith review of NLGR: Utilizing Neighbor Lists for Generative Rerank in Personalized Recommendation Systems." pith.science (2026). https://pith.science/paper/2LYANST3

@misc{pith2026250206097,
  author       = {Pith},
  title        = {Pith review of: NLGR: Utilizing Neighbor Lists for Generative Rerank in Personalized Recommendation Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2LYANST3}},
  note         = {Machine review of arXiv:2502.06097}
}

read the original abstract

Reranking plays a crucial role in modern multi-stage recommender systems by rearranging the initial ranking list. Due to the inherent challenges of combinatorial search spaces, some current research adopts an evaluator-generator paradigm, with a generator generating feasible sequences and an evaluator selecting the best sequence based on the estimated list utility. However, these methods still face two issues. Firstly, due to the goal inconsistency problem between the evaluator and generator, the generator tends to fit the local optimal solution of exposure distribution rather than combinatorial space optimization. Secondly, the strategy of generating target items one by one is difficult to achieve optimality because it ignores the information of subsequent items. To address these issues, we propose a utilizing Neighbor Lists model for Generative Reranking (NLGR), which aims to improve the performance of the generator in the combinatorial space. NLGR follows the evaluator-generator paradigm and improves the generator's training and generating methods. Specifically, we use neighbor lists in combination space to enhance the training process, making the generator perceive the relative scores and find the optimization direction. Furthermore, we propose a novel sampling-based non-autoregressive generation method, which allows the generator to jump flexibly from the current list to any neighbor list. Extensive experiments on public and industrial datasets validate NLGR's effectiveness and we have successfully deployed NLGR on the Meituan food delivery platform.

Figures

Figures reproduced from arXiv: 2502.06097 by the authors.

**Figure 2.** NLGR-E includes two inputs: the exposed list to be evaluated and the user session-level behavior sequence, where each session in the user session-level behavior sequence is the user’s historical exposed list. The evaluation process of NLGR-E is as follows: First, we use an embedding layer to get the embedding of the original input, donated as X ∈ R 𝑚×𝐹 ×𝐷 and M ∈ R 𝐻 ×𝑚×𝐹 ×𝐷 respectively, where 𝐻 represents the numb… view at source ↗

**Figure 2.** The overall architecture of NLGR. where 𝜏 > 0 is a temperature parameter, 𝑛 = −log(−log(𝑢))) represents random noise sampled from the Gumbel distribution, 𝑢 is a uniform distribution between [0, 1]. During backpropagation, the gradient is calculated using Eq. 7. While during forward propagation, the replaced position is 𝑗 = argmax(𝑟 𝑝 𝑗 ). 4.2.2 Candidate Retrieval Unit (CRU). After determining the position 𝑗 to … view at source ↗

Figures from the paper (2 more)

**Figure 3.** Figure 3: The training process of NLGR-G for a candidate list [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]

**Figure 4.** Figure 4: Architecture of the online deployment with NLGR. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 12 canonical work pages

[2]

Qingyao Ai, Keping Bi, Jiafeng Guo, and W Bruce Croft. 2018. Learning a deep listwise context model for ranking refinement. In The 41st international ACM SIGIR conference on research & development in information retrieval . 135–144

work page 2018
[3]

Areej Alsini, Du Q Huynh, and Amitava Datta. 2020. Hit ratio: An evaluation metric for hashtag recommendation. arXiv preprint arXiv:2010.01258 (2020)

arXiv 2020
[4]

Irwan Bello, Sayali Kulkarni, Sagar Jain, Craig Boutilier, Ed Chi, Elad Eban, Xiyang Luo, Alan Mackey, and Ofer Meshi. 2018. Seq2Slate: Re-ranking and slate optimization with RNNs. arXiv preprint arXiv:1810.02019 (2018)

arXiv 2018
[5]

Christopher JC Burges. 2010. From ranknet to lambdarank to lambdamart: An overview. Learning 11, 23-581 (2010), 81

2010
[6]

Zhe Cao, Tao Qin, Tie-Yan Liu, Ming-Feng Tsai, and Hang Li. 2007. Learning to rank: from pairwise approach to listwise approach. In Proceedings of the 24th international conference on Machine learning . 129–136

2007
[7]

Jianxin Chang, Chenbin Zhang, Zhiyi Fu, Xiaoxue Zang, Lin Guan, Jing Lu, Yiqun Hui, Dewei Leng, Yanan Niu, Yang Song, et al. 2023. TWIN: TWo-stage interest network for lifelong user behavior modeling in CTR prediction at kuaishou. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3785–3794

2023
[8]

Chi Chen, Hui Chen, Kangzhi Zhao, Junsheng Zhou, Li He, Hongbo Deng, Jian Xu, Bo Zheng, Yong Zhang, and Chunxiao Xing. 2022. EXTR: Click-Through Rate Prediction with Externalities in E-Commerce Sponsored Search. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 2732–2740

work page 2022
[9]

Qiwei Chen, Changhua Pei, Shanshan Lv, Chao Li, Junfeng Ge, and Wenwu Ou

Show all 40 references

[10]

Heng-Tze Cheng, Levent Koc, Jeremiah Harmsen, Tal Shaked, Tushar Chandra, Hrishi Aradhye, Glen Anderson, Greg Corrado, Wei Chai, Mustafa Ispir, et al
[11]

Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep neural networks for youtube recommendations. In Proceedings of the 10th ACM conference on recommender systems. 191–198

2016
[12]

Yufei Feng, Yu Gong, Fei Sun, Junfeng Ge, and Wenwu Ou. 2021. Revisit recom- mender system in the permutation prospective. arXiv preprint arXiv:2102.12057 (2021)

2021 arXiv
[13]

Yufei Feng, Binbin Hu, Yu Gong, Fei Sun, Qingwen Liu, and Wenwu Ou. 2021. GRN: Generative Rerank Network for Context-wise Recommendation. arXiv preprint arXiv:2104.00860 (2021)

2021 arXiv
[14]

Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. 2017. DeepFM: a factorization-machine based neural network for CTR prediction.arXiv preprint arXiv:1703.04247 (2017)

2017 arXiv
[15]

Iris AM Huijben, Wouter Kool, Max B Paulus, and Ruud JG Van Sloun. 2022. A review of the gumbel-max trick and its extensions for discrete stochasticity in machine learning. IEEE transactions on pattern analysis and machine intelligence 45, 2 (2022), 1353–1371

2022
[16]

Eugene Ie, Vihan Jain, Jing Wang, Sanmit Narvekar, Ritesh Agarwal, Rui Wu, Heng-Tze Cheng, Tushar Chandra, and Craig Boutilier. 2019. SlateQ: A tractable decomposition for reinforcement learning with recommendation sets. (2019)

2019
[17]

Eric Jang, Shixiang Gu, and Ben Poole. 2016. Categorical reparameterization with gumbel-softmax. arXiv preprint arXiv:1611.01144 (2016)

2016 arXiv
[18]

Ray Jiang, Sven Gowal, Timothy A Mann, and Danilo J Rezende. 2018. Beyond greedy ranking: Slate optimization via list-CVAE. arXiv preprint arXiv:1803.01682 (2018)

2018 arXiv
[19]

Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recom- mendation. In 2018 IEEE international conference on data mining (ICDM) . IEEE, 197–206

2018
[20]

Xiao Lin, Xiaokai Chen, Chenyang Wang, Hantao Shu, Linfeng Song, Biao Li, and Peng Jiang. 2024. Discrete conditional diffusion for reranking in recommendation. In Companion Proceedings of the ACM on Web Conference 2024 . 161–169

2024
[21]

Shuchang Liu, Qingpeng Cai, Zhankui He, Bowen Sun, Julian McAuley, Dong Zheng, Peng Jiang, and Kun Gai. 2023. Generative flow network for listwise rec- ommendation. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 1524–1534

2023
[22]

Shuchang Liu, Fei Sun, Yingqiang Ge, Changhua Pei, and Yongfeng Zhang. 2021. Variation control and evaluation for generative slate recommendations. In Pro- ceedings of the Web Conference 2021 . 436–448

2021
[23]

Xiangyu Liu, Chuan Yu, Zhilin Zhang, Zhenzhe Zheng, Yu Rong, Hongtao Lv, Da Huo, Yiqing Wang, Dagui Chen, Jian Xu, et al. 2021. Neural auction: End-to- end learning of auction mechanisms for e-commerce advertising. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge ...

2021
[24]

Medress, Franklin S Cooper, Jim W

Mark F. Medress, Franklin S Cooper, Jim W. Forgie, CC Green, Dennis H. Klatt, Michael H. O’Malley, Edward P Neuburg, Allen Newell, DR Reddy, B Ritea, et al
[25]

Liang Pang, Jun Xu, Qingyao Ai, Yanyan Lan, Xueqi Cheng, and Jirong Wen
[26]

Changhua Pei, Yi Zhang, Yongfeng Zhang, Fei Sun, Xiao Lin, Hanxiao Sun, Jian Wu, Peng Jiang, Junfeng Ge, Wenwu Ou, et al. 2019. Personalized re-ranking for recommendation. In Proceedings of the 13th ACM conference on recommender systems. 3–11

2019
[27]

Qi Pi, Guorui Zhou, Yujing Zhang, Zhe Wang, Lejian Ren, Ying Fan, Xiaoqiang Zhu, and Kun Gai. 2020. Search-based user interest modeling with lifelong sequential behavior data for click-through rate prediction. In Proceedings of the 29th ACM International Conference on Informat...

2020
[28]

Yuxin Ren, Qiya Yang, Yichun Wu, Wei Xu, Yalong Wang, and Zhiqiang Zhang
[29]

Xiaowen Shi, Fan Yang, Ze Wang, Xiaoxu Wu, Muzhi Guan, Guogang Liao, Wang Yongkang, Xingxing Wang, and Dong Wang. 2023. PIER: Permutation-Level Interest-Based End-to-End Re-ranking Framework in E-commerce. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery ...

2023
[30]

A Vaswani. 2017. Attention is all you need. Advances in Neural Information Processing Systems (2017)

2017
[31]

Oriol Vinyals, Meire Fortunato, and Navdeep Jaitly. 2015. Pointer networks. Advances in neural information processing systems 28 (2015)

2015
[32]

Yunjia Xi, Weiwen Liu, Xinyi Dai, Ruiming Tang, Weinan Zhang, Qing Liu, Xi- uqiang He, and Yong Yu. 2021. Context-aware reranking with utility maximization for recommendation. arXiv preprint arXiv:2110.09059 (2021)

2021 arXiv
[33]

Yunjia Xi, Weiwen Liu, Jieming Zhu, Xilong Zhao, Xinyi Dai, Ruiming Tang, Weinan Zhang, Rui Zhang, and Yong Yu. 2022. Multi-Level Interaction Reranking with User Behavior History. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Info...

2022
[34]

Yueqi Xie, Peilin Zhou, and Sunghun Kim. 2022. Decoupled side information fusion for sequential recommendation. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval . 1611–1621

2022
[35]

Guorui Zhou, Xiaoqiang Zhu, Chenru Song, Ying Fan, Han Zhu, Xiao Ma, Yanghui Yan, Junqi Jin, Han Li, and Kun Gai. 2018. Deep interest network for click-through rate prediction. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining ...

2018
[37]

Tao Zhuang, Wenwu Ou, and Zhirong Wang. 2018. Globally optimized mutual influence aware ranking in e-commerce search. arXiv preprint arXiv:1805.08524 (2018)

2018 arXiv
[1977]

Artificial Intelligence 9, 3 (1977), 307–316

Speech understanding systems: Report of a steering committee. Artificial Intelligence 9, 3 (1977), 307–316

1977
[2016]

In Proceedings of the 1st workshop on deep learning for recommender systems

Wide & deep learning for recommender systems. In Proceedings of the 1st workshop on deep learning for recommender systems . 7–10
[2020]

In Proceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval

Setrank: Learning a permutation-invariant ranking model for information retrieval. In Proceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval . 499–508
[2021]

arXiv preprint arXiv:2108.04468 (2021)

End-to-end user behavior retrieval in click-through rateprediction model. arXiv preprint arXiv:2108.04468 (2021)

2021 arXiv
[2024]

In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

Non-autoregressive generative models for reranking recommendation. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 5625–5634

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [2]

Qingyao Ai, Keping Bi, Jiafeng Guo, and W Bruce Croft. 2018. Learning a deep listwise context model for ranking refinement. In The 41st international ACM SIGIR conference on research & development in information retrieval . 135–144

work page 2018

[2] [3]

Areej Alsini, Du Q Huynh, and Amitava Datta. 2020. Hit ratio: An evaluation metric for hashtag recommendation. arXiv preprint arXiv:2010.01258 (2020)

arXiv 2020

[3] [4]

Irwan Bello, Sayali Kulkarni, Sagar Jain, Craig Boutilier, Ed Chi, Elad Eban, Xiyang Luo, Alan Mackey, and Ofer Meshi. 2018. Seq2Slate: Re-ranking and slate optimization with RNNs. arXiv preprint arXiv:1810.02019 (2018)

arXiv 2018

[4] [5]

Christopher JC Burges. 2010. From ranknet to lambdarank to lambdamart: An overview. Learning 11, 23-581 (2010), 81

2010

[5] [6]

Zhe Cao, Tao Qin, Tie-Yan Liu, Ming-Feng Tsai, and Hang Li. 2007. Learning to rank: from pairwise approach to listwise approach. In Proceedings of the 24th international conference on Machine learning . 129–136

2007

[6] [7]

Jianxin Chang, Chenbin Zhang, Zhiyi Fu, Xiaoxue Zang, Lin Guan, Jing Lu, Yiqun Hui, Dewei Leng, Yanan Niu, Yang Song, et al. 2023. TWIN: TWo-stage interest network for lifelong user behavior modeling in CTR prediction at kuaishou. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3785–3794

2023

[7] [8]

Chi Chen, Hui Chen, Kangzhi Zhao, Junsheng Zhou, Li He, Hongbo Deng, Jian Xu, Bo Zheng, Yong Zhang, and Chunxiao Xing. 2022. EXTR: Click-Through Rate Prediction with Externalities in E-Commerce Sponsored Search. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 2732–2740

work page 2022

[8] [9]

Qiwei Chen, Changhua Pei, Shanshan Lv, Chao Li, Junfeng Ge, and Wenwu Ou