Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Large Language Model as Universal Retriever in Industrial-Scale Recommender System

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A single large language model, instructed with a text objective, can retrieve items for many different recommendation goals at industrial scale.

desk verdict Solid industrial systems paper with a real 3% online revenue lift; the sampling theory is thinner than claimed, but the overall approach holds up. read the letter →

arxiv 2502.03041 v2 pith:3LPG2GHT submitted 2025-02-05 cs.IR cs.LG

Junguang Jiang , Yanwen Huang , Bin Liu , Xiaoyu Kong , Xinhang Li , Ziru Xu , Han Zhu , Jian Xu
show 1 more author
Bo Zheng
This is my paper · ORCID
classification cs.IRcs.LG
keywords universalretrievalgenerativelargelanguagemodelsrecommendersystemsmatrixdecompositionmulti-queryrepresentationprobabilisticsamplingcold-startitem
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 claims that one large language model can serve as a universal retriever for an industrial recommender system, handling many retrieval objectives with a single set of weights. Where current systems train separate models for clicks, purchases, scenarios, long-tail items, and query-conditioned retrieval, the proposed Universal Retrieval Model (URM) takes a natural-language description of the objective as part of its input and directly generates a candidate set. To make generative retrieval work with tens of millions of candidates, the paper introduces multi-query representation, a decomposed mapping matrix, and iterative probabilistic sampling. On public datasets URM reports large relative gains over strong baselines, and on industrial data it outperforms task-specific and multi-task baselines on six of nine objectives while improving the platform's core online advertising metric by about 3 percent.

What carries the argument

The load-bearing object is the decomposed candidate map $W = U(V_{\text{dis}}+V_{\text{trans}})^T$ combined with multi-query representation. Instead of learning a full $D \times |C|$ matrix, URM learns a low-rank $U$ and item-side $V$; $V_{\text{dis}}$ preserves per-item discriminability while $V_{\text{trans}}$ transfers to unseen items through serialized attribute text embedded by a general text encoder. The multi-query mechanism appends $M$ learnable tokens to the input and takes the maximum over the resulting score vectors, letting one LLM forward pass express a set of varied user interests. Finally, Algorithm 1 replaces the intractable full softmax with $T$ rounds of probability-weighted sampling over an approximate-nearest-neighbor graph of $W$ rows, reducing matrix-computation cost from roughly 5000 GFLOPs to 2 GFLOPs while approaching exact ranking on the reported industrial tasks.

What would settle it

Run the full $W^T F(u,o)$ scoring over the entire candidate set for a held-out batch of user-objective pairs, then compare its exact top-K against Algorithm 1's output; the claim is falsified if the exact top-K regularly contains items that never enter the sampled neighborhood chain, especially for long-tail items where the neighbor-similarity assumption is weakest.

Watch

Extended reading notes

Core claim

URM's central claim is that retrieval can be reframed as conditional generation, $P(v|u,o)=\mathrm{softmax}(W^T F(u,o))\rvert_v$, where $u$ is a serialized user description, $o$ is a natural-language objective, and $F$ is a large language model used as a feature generator in one forward pass. To give the linear decoding enough expressive power, the model appends $M$ learnable query tokens and uses the maximum of their inner products as the score, so different tokens capture different facets of user interest. To keep the mapping learnable with tens of millions of candidates, $W$ is decomposed into $U(V_{\text{dis}}+V_{\text{trans}})^T$, where $V_{\text{dis}}$ is a fully learnable item embedding for discriminability and $V_{\text{trans}}$ is built from serialized item attributes through a general text embedder for cold-start transferability. To keep inference cheap, full-candidate scoring is replaced by repeated probability-weighted sampling over an approximate-nearest-neighbor graph of items. The paper reports that this pipeline outperforms separately trained expert retrieval models on most industrial objectives and raises advertising revenue by 3.01% in an online A/B test.

Load-bearing premise

The load-bearing premise is that items that are neighbors in the learned item space will also score nearly identically for a given user and objective, so that sampling over a neighbor graph can stand in for scoring all tens of millions of candidates.

Editorial extensions

If this is right

  • A single deployed model can replace per-objective expert retrievers, cutting the training, storage, and serving overhead of multi-channel retrieval.
  • Changing a retrieval goal becomes an input-level operation: an objective such as serendipity or a hybrid like query-by-long-tail is edited in text rather than handled by retraining or rearchitecting.
  • Items never seen in training become retrievable through their attribute text, which matters in catalogs that change minute by minute.
  • Multi-task training on one LLM avoids the seesaw effect better than shared-bottom, MMoE, and PLE baselines, and converges faster with less task-specific data.
  • The system is deployable at industrial latency: one LLM prefill plus a few sampling steps keeps serving within tens of milliseconds, and daily training completes within a 24-hour window on 5% sampled data.

Reading between the lines

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

  • Editorial inference: if prompt-conditioning is as reliable as the reported tables suggest, the same checkpoint could be steered toward new business objectives by writing new objective text, without retraining, for objectives close to the training distribution.
  • Editorial inference: the sampler's neighbor-similarity premise is a natural stress point for long-tail fairness, since rare items have thin neighborhoods; a targeted exact-versus-sampled ranking evaluation on tail items would show whether the reported gains persist as the catalog shifts.
  • Editorial inference: none of the three main components is advertising-specific, so the architecture is a plausible template for other rapidly changing large-scale catalogs, including video, search, and e-commerce, where cold-start transfer and multi-objective retrieval matter equally.
  • Editorial inference: because composing objectives in text changes output statistics, retrieval could become one step inside a larger LLM reasoning loop, with chain-of-thought text injected as context before the candidate set is generated.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes the Universal Retrieval Model (URM), a generative retrieval system in which a fine-tuned LLM (Qwen-7B by default) consumes a serialized user description plus a text-specified retrieval objective and emits a set of item candidates. The technical contributions are: (i) multi-query representation, where M learnable query tokens produce M user representations and the final score is the max over their inner products with item embeddings; (ii) matrix decomposition W = U(Vdis + Vtrans)^T, where Vdis provides item-specific discriminability and Vtrans provides transferability to cold-start items; and (iii) a probabilistic sampling procedure (Algorithm 1) that iteratively expands a random seed set through an ANN index over W to avoid computing scores over the full candidate set. The authors report strong offline results on four public datasets, recall gains over several multi-task baselines on a proprietary industrial dataset with nine retrieval objectives, and an online A/B test on an advertising platform in which revenue improved by 3.01% over a two-week period. The central claim is that a single fine-tuned LLM can serve as a universal retriever that replaces task-specific retrieval models and can be redirected to new objectives by changing the input text.

Significance. If fully validated, the work is practically significant: prompt-conditioned retrieval would remove the need to retrain a separate model for each objective, exploit the LLM's text understanding for zero-shot and hybrid objectives, and provide a path to cold-start item retrieval through Vtrans. The paper has notable strengths: it reports ablations for each main design choice (query token count, matrix decomposition components, sampling steps), tests multiple LLM backbones, includes public dataset comparisons with the strongest baselines, provides substantial implementation detail in the appendices, and candidly discloses limitations and the absence of statistical significance measures. The industrial deployment evidence is valuable but rests on proprietary data, a single uncontrolled online experiment, and a sampling approximation whose theoretical support is a local smoothness bound rather than a coverage guarantee. These gaps need to be closed before the universal-retriever claim can be accepted at face value.

major comments (4)
  1. [Section 3.3, Appendix A.3, Table 4] Algorithm 1 lacks a coverage guarantee, and this gap is load-bearing for the tens-of-milliseconds efficiency claim. Theorem A.1 proves only that nearby columns of W have similar URM scores under a norm cap of B=100; it does not bound the probability that a high-scoring item is contained in the initial random subset Subset(C) or is reachable through the ANN neighbor graph. The size of Subset(C) is unspecified, and the graph expansion or connectivity of NBR(·) is not characterized. Table 4 shows that after T=4 steps the sampled retrieval attains 91.0% of full-matrix recall for one setting, with no error bars and no stratification by item popularity. Because long-tail and cold-start items are exactly those for which W-neighborhoods may be sparse, the online long-tail improvement (+2.23%) and the broader-impact claims in Appendix A.2 are not supported by the sampling analysis. I ask the authors to specify the seed size, report ANN index recall and graph statistics (e.g., reachability of top items from random seeds), stratify Table 4 by popularity buckets, and provide a quantitative comparison of sampled versus exact full-matrix Top-K retrieval on a manageable subset of the candidate set.
  2. [Section 5.5, Table 6, Checklist item 7] The central online claim of 3.01% revenue improvement rests on a single A/B run from April 28 to May 14, 2025, with no confidence intervals, no significance tests, no description of the traffic split, and no reporting of multiple independent periods. The authors themselves acknowledge in the NeurIPS checklist that no statistical significance measures were computed. Under these conditions, the word 'significantly improves' in the abstract is not quantitatively supported. I request confidence intervals or p-values for all four metrics, a description of the experiment design (e.g., percentage of traffic, user-level versus request-level assignment, whether the metric is normalized for seasonality), and ideally a second independent test period or segment-level analysis to confirm the revenue effect.
  3. [Abstract, Section 3.1, Appendix A.1] The abstract's phrase 'any given objective' overstates what the paper demonstrates, and the authors' own limitation statement contradicts the universal claim. Appendix A.1 says that for entirely new objectives it is challenging for URM to adjust its output based on textual input, and that solutions require more training data. The zero-shot evidence in Section 5.4 and Appendix C.4 concerns objectives that are closely related to the training objectives (e.g., RQ learned from CPR data, hybrid objectives, and context queries). I recommend narrowing the claims to 'objectives related to the training distribution' or adding experiments on genuinely distant objectives (e.g., a task with different output semantics or a domain outside the training scenarios) to substantiate the universality claim.
  4. [Section 5.2, Appendix B.3, Table 2] The baseline comparison in Table 2 is confounded by model scale and training data. URM uses a 7B-parameter LLM (or 1.8B/13B in Appendix C.2) fine-tuned on 5% of aggregated data (Appendix B.3), while the two-tower, Transformer, and Attention-DNN baselines are orders of magnitude smaller. The paper does not state whether the baselines were trained on the same 5% sample or on the full data, nor does it report compute-matched comparisons. To make the claim that URM 'outperforms expert models' attributable to the framework rather than to model capacity, the authors should report baselines under matched data and a comparable compute budget, or at least provide an ablation in which a non-LLM model of comparable capacity is trained with the same objective text conditioning and multi-query architecture.
minor comments (5)
  1. [Appendix A.3] In Theorem A.1 and the surrounding text, the norm on the left-hand side of Eq. (4) should be an absolute value because max(W^T_v1 F(u,o)) is a scalar; the overloading of the norm symbol is confusing. Also, Eq. (3) uses F(u,o) on both sides of the definition, which makes the norm-cap appear circular; please use a different symbol for the scaled representation.
  2. [Section 3.3] The complexity expression O(M H(D + T K× max(|NBR(·)|)) has unbalanced parentheses and undefined bracket placement. Please state the per-step cost of Algorithm 1 explicitly, including the cost of computing NBR(·) lookups and the size of the initial subset.
  3. [Table 4] The 'Precision' metric is defined only informally in the text after Table 4 as a ratio of R@1000 values; please define it precisely in the main text, state the candidate-set size and K, and report the seed subset size used in Algorithm 1.
  4. [Section 5.3, Table 17] The claim in Section 3.1 that 'the max function works better than linear combinations' is not supported by the presented ablations, which vary the number of tokens and the embedding dimension but do not compare combine operators. Please add an ablation comparing max, mean, sum, and a learned combination.
  5. [Appendix B.3] The fact that URM is trained on only 5% of aggregated data is disclosed in Appendix B.3 but should be stated in Section 5.2 alongside the industrial results, since it is essential for interpreting the offline recall numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: URM's derivation chain is a standard supervised learning pipeline whose components are validated on held-out data and external benchmarks, not by construction or by load-bearing self-citation.

full rationale

The paper's derivation chain is self-contained against external evidence. The core objective P(v|u,o) = softmax(max(W^T F(u,o))) is a standard discriminative model trained with NCE and evaluated on held-out test sets and public benchmarks against a wide range of external baselines (SASRec, BERT4Rec, IDGenRec, COBRA, etc.), so the main performance claims do not reduce to the training inputs. The matrix decomposition W = U(Vdis+Vtrans)^T is a low-rank parameterization whose benefits are demonstrated by ablations rather than assumed by definition. Probabilistic sampling (Algorithm 1) is explicitly an approximation: Table 4 measures its precision against full-matrix ranking, and Theorem A.1 in Appendix A.3 only establishes a local Lipschitz bound on score differences under the stated neighbor-similarity assumption. That theorem restates the algorithm's working assumption in mathematical form; it is not a definitional equivalence between the sampled ranking and the full ranking. The absence of a coverage guarantee for long-tail items is a correctness/robustness concern, but not a circularity. Self-citations to prior Alibaba tree-based models and ForkMerge appear only as background or baselines and are not load-bearing for URM's claims. The abstract's phrase 'arbitrary given objective' is also tempered by the paper's own limitation in Appendix A.1, which acknowledges difficulty with entirely new objectives; this is an overclaim relative to the experiments, not a circular step. No equation is defined in terms of the result it is said to predict, and no fitted quantity is renamed as a prediction. The paper's central contributions are empirically established against independent benchmarks and held-out online traffic, so the circularity score is 0.

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

No new physical or conceptual entities are postulated. Multi-query tokens and the decomposed matrices are learned parameters within the model architecture. The paper relies on domain assumptions about neighborhood similarity, the value of multi-query max, and transferability of fine-tuned LLMs; these are empirical design choices rather than independently verified facts.

free parameters (6)
  • sampling temperature tau = 0.07
    Used in Algorithm 1 softmax to control sampling sharpness; set by hand and no sensitivity analysis is reported.
  • norm bound B = 100
    Bound on the user representation norm in Theorem A.1; chosen by hand to trade expressiveness against neighbor score consistency.
  • sampling steps T = 4
    Number of iterative sampling rounds in Algorithm 1; Table 4 shows the ratio R@1000 with sampling to R@1000 without sampling reaches 91.0% at T=4, which is used in production.
  • query token count M = 128
    Number of learnable query tokens producing M user representations; Figure 3 shows performance improves with M, so a large M is used.
  • matrix rank H = 128
    Rank of the decomposed matrices U and V in W = U(Vdis + Vtrans)^T; this is a hyperparameter and the dimension of the user/item representations in the logit computation.
  • embedding dimension D = 4096
    Hidden dimension of the LLM and the mapping MLP; tied to the Qwen-7B backbone and not independently tuned.
assumptions (5)
  • domain assumption Items close in the W matrix will receive similar scores P(v|u,o) from URM
    Explicitly stated in Section 3.3 as the basis for Algorithm 1. Theorem A.1 proves only a Lipschitz bound under a norm bound B, not that the sampled set converges to the full-matrix ranking.
  • ad hoc to paper A linear combination or max over multiple learned representations can approximate complex user-item interactions
    Section 3.1 motivates multi-query representation by function approximation theory, but the specific choice of max over linear combination is justified only by the paper's own ablations (Figure 3 and Appendix C.5).
  • domain assumption Pre-trained LLM knowledge is retained and transferable after fully fine-tuning on retrieval data sampled at 5%
    Section 3.1 and Appendix B.3 use a 5% sample of industrial data and a reduced learning rate for pretrained layers; no measurement of retained world knowledge is provided.
  • domain assumption The decomposed mapping W = U(Vdis + Vtrans)^T separates discriminative and transferable item representations
    Section 3.2 asserts the decomposition improves learnability, discriminability, and transferability; evidence is the ablation in Table 3 rather than a formal guarantee.
  • ad hoc to paper The max over M query outputs, rather than a higher-dimensional single output, is the correct way to use model capacity
    Appendix C.5 Table 17 shows a split 4096-d single query performs worse than 32x128-d multi-query, but this is an empirical comparison on two tasks, not a derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large Language Model as Universal Retriever in Industrial-Scale Recommender System." pith.science (2026). https://pith.science/paper/3LPG2GHT

@misc{pith2026250203041,
  author       = {Pith},
  title        = {Pith review of: Large Language Model as Universal Retriever in Industrial-Scale Recommender System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3LPG2GHT}},
  note         = {Machine review of arXiv:2502.03041}
}
abstract

In real-world recommender systems, different retrieval objectives are typically addressed using task-specific datasets with carefully designed model architectures. We demonstrate that Large Language Models (LLMs) can function as universal retrievers, capable of handling multiple objectives within a generative retrieval framework. To model complex user-item relationships within generative retrieval, we propose multi-query representation. To address the challenge of extremely large candidate sets in industrial recommender systems, we introduce matrix decomposition to boost model learnability, discriminability, and transferability, and we incorporate probabilistic sampling to reduce computation costs. Finally, our Universal Retrieval Model (URM) can adaptively generate a set from tens of millions of candidates based on arbitrary given objective while keeping the latency within tens of milliseconds. Applied to industrial-scale data, URM outperforms expert models elaborately designed for different retrieval objectives on offline experiments and significantly improves the core metric of online advertising platform by $3\%$.

Figures

Figures reproduced from arXiv: 2502.03041 by the authors.

Figure 1
Figure 1. URM architecture. The input sequence consists of user description [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Online Serving System [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. The effect of query token number M. 1 2 4 8 16 32 64 128 M 0.18 0.20 0.22 0.24 0.26 CPR R@1000 [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Performance on unseen queries. Zero-shot Learning. We find that URM effectively adapts to hybrid objectives, such as long-tail item re￾trieval with a specific query. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: The effectiveness of multi-task learning. [PITH_FULL_IMAGE:figures/full_fig_p022_5.png]
Figure 6
Figure 6. Figure 6: Zero-shot task transfer performance from the CPR task to the RQ task. [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]
Figure 7
Figure 7. Figure 7: Zero-shot task transfer to a new context. Christmas is a festive celebration, and heating is [PITH_FULL_IMAGE:figures/full_fig_p024_7.png]
Figure 8
Figure 8. Figure 8: Visualization of user representations from different query tokens. Each token captures [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: Activation proportion of different query tokens in CPR and RQ tasks. [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: Visualization of different item representations (case 1). [PITH_FULL_IMAGE:figures/full_fig_p026_10.png]
Figure 11
Figure 11. Figure 11: Visualization of different item representations (case 2). [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]
Figure 12
Figure 12. Figure 12: Visualization of different item representations (case 3). [PITH_FULL_IMAGE:figures/full_fig_p027_12.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. OneSug: The Unified End-to-End Generative Framework for E-commerce Query Suggestion

    cs.IR 2025-06 conditional novelty 6.0 of 10

    A unified encoder-decoder model with prefix representation enhancement and reward-weighted DPO outperforms Kuaishou's online multi-stage query suggestion system in offline and live A/B evaluations.

  2. GR-LLMs: Recent Advances in Generative Recommendation Based on Large Language Models

    cs.IR 2025-07 unverdicted novelty 3.0 of 10

    A survey of LLM-based generative recommendation systems, covering application settings, training pipelines, industrial deployment challenges, and future directions.

Reference graph

Works this paper leans on

120 extracted references · 49 canonical work pages · cited by 2 Pith papers

  1. [1]

    Gpt-4 technical report

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023

    arXiv 2023

  2. [2]

    Qwen technical report

    Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, Binyuan Hui, Luo Ji, Mei Li, Junyang Lin, Runji Lin, Dayiheng Liu, Gao Liu, Chengqiang Lu, Keming Lu, Jianxin Ma, Rui Men, Xingzhang Ren, Xuancheng Ren, Chuanqi Tan, Sinan Tan, Jianhong Tu, Peng Wang, Shijie Wang, Wei Wang, Shengguang Wu, Benfeng X...

    arXiv 2023

  3. [3]

    Chime: A compressive framework for holistic interest modeling

    Yong Bai, Rui Xiang, Kaiyuan Li, Yongxiang Tang, Yanhua Cheng, Xialong Liu, Peng Jiang, and Kun Gai. Chime: A compressive framework for holistic interest modeling. CoRR, abs/2504.06780, 2025

    arXiv 2025

  4. [4]

    A bi-step grounding paradigm for large language models in recommendation systems

    Keqin Bao, Jizhi Zhang, Wenjie Wang, Yang Zhang, Zhengyi Yang, Yancheng Luo, Chong Chen, Fuli Feng, and Qi Tian. A bi-step grounding paradigm for large language models in recommendation systems. arXiv preprint arXiv:2308.08434, 2023

    arXiv 2023

  5. [5]

    Tallrec: An effective and efficient tuning framework to align large language model with recommendation

    Keqin Bao, Jizhi Zhang, Yang Zhang, Wenjie Wang, Fuli Feng, and Xiangnan He. Tallrec: An effective and efficient tuning framework to align large language model with recommendation. In RecSys, 2023

    2023

  6. [6]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agar- wal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Lit...

    2020

  7. [7]

    Multitask learning

    Rich Caruana. Multitask learning. Machine Learning, 28:41–75, 1997

    1997

  8. [8]

    Hllm: Enhancing sequential recom- mendations via hierarchical large language models for item and user modeling

    Junyi Chen, Lu Chi, Bingyue Peng, and Zehuan Yuan. Hllm: Enhancing sequential recom- mendations via hierarchical large language models for item and user modeling. arXiv preprint arXiv:2409.12740, 2024

    arXiv 2024

Show all 120 references
  1. [9]

    Approximate nearest neighbor search under neural similarity metric for large-scale recommendation

    Rihan Chen, Bin Liu, Han Zhu, Yaoxuan Wang, Qi Li, Buting Ma, Qingbo Hua, Jun Jiang, Yunlong Xu, Hongbo Deng, et al. Approximate nearest neighbor search under neural similarity metric for large-scale recommendation. In CIKM, 2022

    2022

  2. [10]

    Gradnorm: Gradi- ent normalization for adaptive loss balancing in deep multitask networks

    Zhao Chen, Vijay Badrinarayanan, Chen-Yu Lee, and Andrew Rabinovich. Gradnorm: Gradi- ent normalization for adaptive loss balancing in deep multitask networks. In ICML, 2018

    2018

  3. [11]

    Uncovering chatgpt’s capabilities in recommender systems

    Sunhao Dai, Ninglu Shao, Haiyuan Zhao, Weijie Yu, Zihua Si, Chen Xu, ZhongXiang Sun, Xiao Zhang, and Jun Xu. Uncovering chatgpt’s capabilities in recommender systems. Pro- ceedings of the 17th ACM Conference on Recommender Systems, 2023

    2023

  4. [12]

    FlashAttention-2: Faster attention with better parallelism and work partitioning

    Tri Dao. FlashAttention-2: Faster attention with better parallelism and work partitioning. In International Conference on Learning Representations (ICLR), 2024

    2024

  5. [13]

    Fu, Stefano Ermon, Atri Rudra, and Christopher Ré

    Tri Dao, Daniel Y . Fu, Stefano Ermon, Atri Rudra, and Christopher Ré. FlashAttention: Fast and memory-efficient exact attention with IO-awareness. In Advances in Neural Information Processing Systems (NeurIPS), 2022

    2022

  6. [14]

    Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model, 2024

    DeepSeek-AI. Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model, 2024. 10

    2024

  7. [15]

    Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment

    Jiaxin Deng, Shiyao Wang, Kuo Cai, Lejian Ren, Qigen Hu, Weifeng Ding, Qiang Luo, and Guorui Zhou. Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment. CoRR, abs/2502.18965, 2025

    2025 arXiv

  8. [16]

    Deep retrieval: Learning a retrievable structure for large-scale recommendations, 2021

    Weihao Gao, Xiangjun Fan, Chong Wang, Jiankai Sun, Kai Jia, Wenzhi Xiao, Ruofan Ding, Xingyan Bin, Hui Yang, and Xiaobing Liu. Deep retrieval: Learning a retrievable structure for large-scale recommendations, 2021

    2021

  9. [17]

    Recommendation as language processing (RLP): A unified pretrain, personalized prompt & predict paradigm (P5)

    Shijie Geng, Shuchang Liu, Zuohui Fu, Yingqiang Ge, and Yongfeng Zhang. Recommendation as language processing (RLP): A unified pretrain, personalized prompt & predict paradigm (P5). In RecSys, 2022

    2022

  10. [18]

    Recommendation as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5)

    Shijie Geng, Shuchang Liu, Zuohui Fu, Yingqiang Ge, and Yongfeng Zhang. Recommendation as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5). In Proceedings of the Sixteenth ACM Conference on Recommender Systems, 2022

    2022

  11. [19]

    Noise-contrastive estimation: A new estimation principle for unnormalized statistical models

    Michael Gutmann and Aapo Hyvärinen. Noise-contrastive estimation: A new estimation principle for unnormalized statistical models. In Proceedings of the thirteenth international conference on artificial intelligence and statistics , pages 297–304. JMLR Workshop and Conference P...

    2010

  12. [20]

    HANDSCOMB, D.F

    D.C. HANDSCOMB, D.F. MAYERS, and M.J.D. POWELL. Chapter 7 - the general theory of linear approximation. In D.C. HANDSCOMB, editor, Methods of Numerical Approximation, pages 61–71. Pergamon, 1966

    1966

  13. [21]

    Scaling laws for transfer

    Danny Hernandez, Jared Kaplan, Tom Henighan, and Sam McCandlish. Scaling laws for transfer. arXiv preprint arXiv:2102.01293, 2021

    2021 arXiv

  14. [22]

    Multitask-centernet (mcn): Efficient and diverse multitask learning using an anchor free approach

    Falk Heuer, Sven Mantowsky, Saqib Bukhari, and Georg Schneider. Multitask-centernet (mcn): Efficient and diverse multitask learning using an anchor free approach. In ICCV, 2021

    2021

  15. [23]

    Session-based recommendations with recurrent neural networks

    Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk. Session-based recommendations with recurrent neural networks. In ICLR, 2016

    2016

  16. [24]

    The curious case of neural text degeneration

    Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. The curious case of neural text degeneration. In ICLR, 2020

    2020

  17. [25]

    Towards universal sequence representation learning for recommender systems

    Yupeng Hou, Shanlei Mu, Wayne Xin Zhao, Yaliang Li, Bolin Ding, and Ji-Rong Wen. Towards universal sequence representation learning for recommender systems. In KDD, 2022

    2022

  18. [26]

    Embedding-based retrieval in facebook search

    Jui-Ting Huang, Ashish Sharma, Shuying Sun, Li Xia, David Zhang, Philip Pronin, Janani Padmanabhan, Giuseppe Ottaviano, and Linjun Yang. Embedding-based retrieval in facebook search. In KDD. ACM, 2020

    2020

  19. [27]

    A comprehensive survey on retrieval methods in recommender systems, 2024

    Junjie Huang, Jizheng Chen, Jianghao Lin, Jiarui Qin, Ziming Feng, Weinan Zhang, and Yong Yu. A comprehensive survey on retrieval methods in recommender systems, 2024

    2024

  20. [28]

    Unleashing the potential of multi-channel fusion in retrieval for personalized recommendations

    Junjie Huang, Jiarui Qin, Jianghao Lin, Ziming Feng, Weinan Zhang, and Yong Yu. Unleashing the potential of multi-channel fusion in retrieval for personalized recommendations. In WWW, 2025

    2025

  21. [29]

    Knowledge adaptation from large language model to recommendation for practical industrial application

    Jian Jia, Yipei Wang, Yan Li, Honggang Chen, Xuehan Bai, Zhaocheng Liu, Jian Liang, Quan Chen, Han Li, Peng Jiang, and Kun Gai. Knowledge adaptation from large language model to recommendation for practical industrial application. ArXiv, abs/2405.03988, 2024

    2024 arXiv

  22. [30]

    Forkmerge: Mitigating negative transfer in auxiliary-task learning

    Junguang Jiang, Baixu Chen, Junwei Pan, Ximei Wang, Liu Dapeng, Jie Jiang, and Mingsheng Long. Forkmerge: Mitigating negative transfer in auxiliary-task learning. In NeurIPS, 2023

    2023

  23. [31]

    Adaptive domain interest network for multi-domain recommendation

    Yuchen Jiang, Qi Li, Han Zhu, Jinbei Yu, Jin Li, Ziru Xu, Huihui Dong, and Bo Zheng. Adaptive domain interest network for multi-domain recommendation. In CIKM, 2022

    2022

  24. [32]

    Wang-Cheng Kang and Julian J. McAuley. Self-attentive sequential recommendation. In ICDM, 2018. 11

    2018

  25. [33]

    Scaling laws for neural language models

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361, 2020

    2001 arXiv

  26. [34]

    Multi-task learning using uncertainty to weigh losses for scene geometry and semantics

    Alex Kendall, Yarin Gal, and Roberto Cipolla. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In CVPR, 2018

    2018

  27. [35]

    Ubernet: Training a ‘universal’ convolutional neural network for low-, mid-, and high-level vision using diverse datasets and limited memory

    Iasonas Kokkinos. Ubernet: Training a ‘universal’ convolutional neural network for low-, mid-, and high-level vision using diverse datasets and limited memory. In CVPR, 2017

    2017

  28. [36]

    Customizing language models with instance-wise lora for sequential recommendation

    Xiaoyu Kong, Jiancan Wu, An Zhang, Leheng Sheng, Hui Lin, Xiang Wang, and Xiangnan He. Customizing language models with instance-wise lora for sequential recommendation. In NeurIPS, 2024

    2024

  29. [37]

    Rethinking serendipity in recommender systems

    Denis Kotkov, Alan Medlar, and Dorota Glowacka. Rethinking serendipity in recommender systems. In SIGIR, 2023

    2023

  30. [38]

    Gonzalez, Hao Zhang, and Ion Stoica

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems...

    2023

  31. [39]

    Bbqrec: Behavior-bind quantization for multi-modal sequential recommendation

    Kaiyuan Li, Rui Xiang, Yong Bai, Yongxiang Tang, Yanhua Cheng, Xialong Liu, Peng Jiang, and Kun Gai. Bbqrec: Behavior-bind quantization for multi-modal sequential recommendation. CoRR, abs/2504.06636, 2025

    2025 arXiv

  32. [40]

    Andersen, Jun Woo Park, Alexander J

    Mu Li, David G. Andersen, Jun Woo Park, Alexander J. Smola, Amr Ahmed, Vanja Josifovski, James Long, Eugene J. Shekita, and Bor-Yiing Su. Scaling distributed machine learning with the parameter server. In Proceedings of the 11th USENIX Conference on Operating Systems Design an...

    2014

  33. [41]

    Exploring the upper limits of text-based collaborative filtering using large language models: Discoveries and insights

    Ruyu Li, Wenhao Deng, Yu Cheng, Zheng Yuan, Jiaqi Zhang, and Fajie Yuan. Exploring the upper limits of text-based collaborative filtering using large language models: Discoveries and insights. CoRR, abs/2305.11700, 2023

    2023

  34. [42]

    E4srec: An elegant effective efficient extensible solution of large language models for sequential recommendation

    Xinhang Li, Chong Chen, Xiangyu Zhao, Yong Zhang, and Chunxiao Xing. E4srec: An elegant effective efficient extensible solution of large language models for sequential recommendation. CoRR, abs/2312.02443, 2023

    2023 arXiv

  35. [43]

    Towards general text embeddings with multi-stage contrastive learning

    Zehan Li, Xin Zhang, Yanzhao Zhang, Dingkun Long, Pengjun Xie, and Meishan Zhang. Towards general text embeddings with multi-stage contrastive learning. arXiv preprint arXiv:2308.03281, 2023

    2023 arXiv

  36. [44]

    Llara: Aligning large language models with sequential recommenders

    Jiayi Liao, Sihang Li, Zhengyi Yang, Jiancan Wu, Yancheng Yuan, and Xiang Wang. Llara: Aligning large language models with sequential recommenders. CoRR, abs/2312.02445, 2023

    2023 arXiv

  37. [45]

    Unified semantic and ID representation learning for deep recommenders

    Guanyu Lin, Zhigang Hua, Tao Feng, Shuang Yang, Bo Long, and Jiaxuan You. Unified semantic and ID representation learning for deep recommenders. CoRR, abs/2502.16474, 2025

    2025 arXiv

  38. [46]

    Kuaiformer: Transformer-based retrieval at kuaishou, 2024

    Chi Liu, Jiangxia Cao, Rui Huang, Kai Zheng, Qiang Luo, Kun Gai, and Guorui Zhou. Kuaiformer: Transformer-based retrieval at kuaishou, 2024

    2024

  39. [47]

    A unified search and recommendation framework based on multi-scenario learning for ranking in e-commerce

    Jinhan Liu, Qiyu Chen, Junjie Xu, Junjie Li, Baoli Li, and Sulong Xu. A unified search and recommendation framework based on multi-scenario learning for ranking in e-commerce. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Informat...

    2024

  40. [48]

    Is chatgpt a good recommender? A preliminary study

    Junling Liu, Chao Liu, Renjie Lv, Kang Zhou, and Yan Zhang. Is chatgpt a good recommender? A preliminary study. CoRR, abs/2304.10149, 2023

    2023 arXiv

  41. [49]

    End-to-end multi-task learning with attention

    Shikun Liu, Edward Johns, and Andrew J Davison. End-to-end multi-task learning with attention. In CVPR, 2019. 12

    2019

  42. [50]

    Long-tail session-based recommendation

    Siyi Liu and Yujia Zheng. Long-tail session-based recommendation. In Proceedings of the 14th ACM Conference on Recommender Systems, 2020

    2020

  43. [51]

    Mingsheng Long, Yue Cao, Jianmin Wang, and Michael I. Jordan. Learning transferable features with deep adaptation networks. In ICML, 2015

    2015

  44. [52]

    QARM: quantitative alignment multi-modal recommendation at kuaishou

    Xinchen Luo, Jiangxia Cao, Tianyu Sun, Jinkai Yu, Rui Huang, Wei Yuan, Hezheng Lin, Yichen Zheng, Shiyao Wang, Qigen Hu, Changqing Qiu, Jiaqi Zhang, Xu Zhang, Zhiheng Yan, Jingming Zhang, Simin Zhang, Mingxing Wen, Zhaojie Liu, Kun Gai, and Guorui Zhou. QARM: quantitative alig...

    2024 arXiv

  45. [53]

    Hierarchical gating networks for sequential recommen- dation

    Chen Ma, Peng Kang, and Xue Liu. Hierarchical gating networks for sequential recommen- dation. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2019, Anchorage, AK, USA, August 4-8, 2019, 2019

    2019

  46. [54]

    Modeling task relationships in multi-task learning with multi-gate mixture-of-experts

    Jiaqi Ma, Zhe Zhao, Xinyang Yi, Jilin Chen, Lichan Hong, and Ed H Chi. Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. In SIGKDD, 2018

    2018

  47. [55]

    Entire space multi-task model: An effective approach for estimating post-click conversion rate

    Xiao Ma, Liqin Zhao, Guan Huang, Zhi Wang, Zelin Hu, Xiaoqiang Zhu, and Kun Gai. Entire space multi-task model: An effective approach for estimating post-click conversion rate. In SIGIR, 2018

    2018

  48. [56]

    Efficient and robust approximate nearest neighbor search using hierarchical navigable small world graphs

    Yu Malkov and Dmitry Yashunin. Efficient and robust approximate nearest neighbor search using hierarchical navigable small world graphs. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016

    2016

  49. [57]

    Attentive single-tasking of multiple tasks

    Kevis-Kokitsi Maninis, Ilija Radosavovic, and Iasonas Kokkinos. Attentive single-tasking of multiple tasks. In CVPR, 2019

    2019

  50. [58]

    McAuley, Christopher Targett, Qinfeng Shi, and Anton van den Hengel

    Julian J. McAuley, Christopher Targett, Qinfeng Shi, and Anton van den Hengel. Image-based recommendations on styles and substitutes. In SIGIR, 2015

    2015

  51. [59]

    Cross-stitch networks for multi-task learning

    Ishan Misra, Abhinav Shrivastava, Abhinav Gupta, and Martial Hebert. Cross-stitch networks for multi-task learning. In CVPR, 2016

    2016

  52. [60]

    Improving language understanding by generative pre-training

    Alec Radford, Karthik Narasimhan, Tim Salimans, and Ilya Sutskever. Improving language understanding by generative pre-training. Technical report, OpenAI, 2018

    2018

  53. [61]

    Lan- guage models are unsupervised multitask learners

    Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Lan- guage models are unsupervised multitask learners. Technical report, OpenAI, 2019

    2019

  54. [62]

    Tran, Jonah Samost, Maciej Kula, Ed H

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Keshavan, Trung Vu, Lukasz Heidt, Lichan Hong, Yi Tay, Vinh Q. Tran, Jonah Samost, Maciej Kula, Ed H. Chi, and Maheswaran Sathiamoorthy. Recommender systems with generative retrieval. In Proceedings of the 37th Internatio...

    2023

  55. [63]

    Victor Sanh, Albert Webson, Colin Raffel, Stephen H. Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal V . Nayak, ...

    2022

  56. [64]

    Roformer: Enhanced transformer with rotary position embedding, 2023

    Jianlin Su, Yu Lu, Shengfeng Pan, Ahmed Murtadha, Bo Wen, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding, 2023

    2023

  57. [65]

    Bert4rec: Sequential recommendation with bidirectional encoder representations from transformer

    Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang. Bert4rec: Sequential recommendation with bidirectional encoder representations from transformer. In CIKM, 2019. 13

    2019

  58. [66]

    Learning to tokenize for generative retrieval

    Weiwei Sun, Lingyong Yan, Zheng Chen, Shuaiqiang Wang, Haichao Zhu, Pengjie Ren, Zhumin Chen, Dawei Yin, Maarten de Rijke, and Zhaochun Ren. Learning to tokenize for generative retrieval. In NeurIPS, 2023

    2023

  59. [67]

    Idgen- rec: Llm-recsys alignment with textual id learning

    Juntao Tan, Shuyuan Xu, Wenyue Hua, Yingqiang Ge, Zelong Li, and Yongfeng Zhang. Idgen- rec: Llm-recsys alignment with textual id learning. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2024

    2024

  60. [68]

    Progressive layered extraction (ple): A novel multi-task learning (mtl) model for personalized recommendations

    Hongyan Tang, Junning Liu, Ming Zhao, and Xudong Gong. Progressive layered extraction (ple): A novel multi-task learning (mtl) model for personalized recommendations. In RecSys, 2020

    2020

  61. [69]

    Personalized top-n sequential recommendation via convolutional sequence embedding

    Jiaxi Tang and Ke Wang. Personalized top-n sequential recommendation via convolutional sequence embedding. In WSDM, 2018

    2018

  62. [70]

    Llama: Open and efficient foundation language models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timo- thée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurélien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. Llama: Open and efficient foundation langua...

    2023 arXiv

  63. [71]

    Llama 2: Open foundation and fine-tuned chat models

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton-Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, W...

    2023 arXiv

  64. [72]

    Vanchinathan, Isidor Nikolic, Fabio De Bona, and Andreas Krause

    Hastagiri P. Vanchinathan, Isidor Nikolic, Fabio De Bona, and Andreas Krause. Explore- exploit in top-n recommender systems via gaussian processes. In Proceedings of the 8th ACM Conference on Recommender Systems, 2014

    2014

  65. [73]

    Jose, Chenyun Yu, Beibei Kong, Xiangnan He, Zhijin Wang, Bo Hu, and Zang Li

    Jie Wang, Fajie Yuan, Mingyue Cheng, Joemon M. Jose, Chenyun Yu, Beibei Kong, Xiangnan He, Zhijin Wang, Bo Hu, and Zang Li. Transrec: Learning transferable recommendation from mixture-of-modality feedback. In APWeb-WAIM, 2024

    2024

  66. [74]

    Learnable item tokenization for generative recommendation

    Wenjie Wang, Honghui Bao, Xinyu Lin, Jizhi Zhang, Yongqi Li, Fuli Feng, See-Kiong Ng, and Tat-Seng Chua. Learnable item tokenization for generative recommendation. In CIKM, pages 2400–2409, 2024

    2024

  67. [75]

    Self-consistency improves chain of thought reasoning in language models

    Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. In ICLR, 2023

    2023

  68. [76]

    EAGER: two-stream generative recommender with behavior-semantic collaboration

    Ye Wang, Jiahao Xun, Minjie Hong, Jieming Zhu, Tao Jin, Wang Lin, Haoyuan Li, Linjun Li, Yan Xia, Zhou Zhao, and Zhenhua Dong. EAGER: two-stream generative recommender with behavior-semantic collaboration. In SIGKDD, pages 3245–3254, 2024

    2024

  69. [77]

    Chain-of-thought prompting elicits reasoning in large language models

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. In NeurIPS, 2022

    2022

  70. [78]

    Unifying generative and dense retrieval for sequential recommendation, 2024

    Liu Yang, Fabian Paischer, Kaveh Hassani, Jiacheng Li, Shuai Shao, Zhang Gabriel Li, Yun He, Xue Feng, Nima Noorshams, Sem Park, Bo Long, Robert D Nowak, Xiaoli Gao, and Hamid Eghbalzadeh. Unifying generative and dense retrieval for sequential recommendation, 2024. 14

    2024

  71. [79]

    Sparse meets dense: Unified generative recommendations with cascaded sparse-dense representations

    Yuhao Yang, Zhi Ji, Zhaopeng Li, Yi Li, Zhonglin Mo, Yue Ding, Kai Chen, Zijian Zhang, Jie Li, Shuanglong Li, and Lin Liu. Sparse meets dense: Unified generative recommendations with cascaded sparse-dense representations. arXiv preprint arXiv:2503.02453, 2025

    2025 arXiv

  72. [80]

    Harnessing multimodal large language models for multimodal sequential recommendation

    Yuyang Ye, Zhi Zheng, Yishan Shen, Tianshu Wang, Hengruo Zhang, Peijun Zhu, Runlong Yu, Kai Zhang, and Hui Xiong. Harnessing multimodal large language models for multimodal sequential recommendation. arXiv preprint arXiv:2408.09698, 2024

    2024 arXiv

  73. [81]

    Yelp open dataset

    Yelp. Yelp open dataset. https://business.yelp.com/data/resources/ open-dataset/, 2025. Accessed: 2025-04-01

    2025

  74. [82]

    Gradient surgery for multi-task learning

    Tianhe Yu, Saurabh Kumar, Abhishek Gupta, Sergey Levine, Karol Hausman, and Chelsea Finn. Gradient surgery for multi-task learning. In NeurIPS, 2020

    2020

  75. [83]

    Where to go next for recommender systems? ID- vs

    Zheng Yuan, Fajie Yuan, Yu Song, Youhua Li, Junchen Fu, Fei Yang, Yunzhu Pan, and Yongxin Ni. Where to go next for recommender systems? ID- vs. modality-based recommender models revisited. In SIGIR, 2023

    2023

  76. [84]

    Shen, Leonidas J

    Amir Roshan Zamir, Alexander Sax, William B. Shen, Leonidas J. Guibas, Jitendra Malik, and Silvio Savarese. Taskonomy: Disentangling task transfer learning. In CVPR, 2018

    2018

  77. [85]

    Revisit- ing neural retrieval on accelerators

    Jiaqi Zhai, Zhaojie Gong, Yueming Wang, Xiao Sun, Zheng Yan, Fu Li, and Xing Liu. Revisit- ing neural retrieval on accelerators. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023

    2023

  78. [86]

    Actions speak louder than words: Trillion-parameter sequential transducers for generative recommendations

    Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhaojie Gong, Fangda Gu, Jiayuan He, Yinghai Lu, and Yu Shi. Actions speak louder than words: Trillion-parameter sequential transducers for generative recommendations. In ICML, 2024

    2024

  79. [87]

    Root mean square layer normalization

    Biao Zhang and Rico Sennrich. Root mean square layer normalization. In NeurIPS, 2019

    2019

  80. [88]

    Rec- ommendation as instruction following: A large language model empowered recommendation approach

    Junjie Zhang, Ruobing Xie, Yupeng Hou, Wayne Xin Zhao, Leyu Lin, and Ji-Rong Wen. Rec- ommendation as instruction following: A large language model empowered recommendation approach. CoRR, abs/2305.07001, 2023

    2023 arXiv

  81. [89]

    Learning id-free item representation with token crossing for multimodal recommenda- tion

    Kangning Zhang, Jiarui Jin, Yingjie Qin, Ruilong Su, Jianghao Lin, Yong Yu, and Weinan Zhang. Learning id-free item representation with token crossing for multimodal recommenda- tion. CoRR, abs/2410.19276, 2024

    2024 arXiv

  82. [90]

    Towards scalable semantic representation for recommendation

    Taolin Zhang, Junwei Pan, Jinpeng Wang, Yaohua Zha, Tao Dai, Bin Chen, Ruisheng Luo, Xiaoxiang Deng, Yuan Wang, Ming Yue, Jie Jiang, and Shu-Tao Xia. Towards scalable semantic representation for recommendation. CoRR, abs/2410.09560, 2024

    arXiv

  83. [91]

    Sheng, Jiajie Xu, Deqing Wang, Guanfeng Liu, and Xiaofang Zhou

    Tingting Zhang, Pengpeng Zhao, Yanchi Liu, Victor S. Sheng, Jiajie Xu, Deqing Wang, Guanfeng Liu, and Xiaofang Zhou. Feature-level deeper self-attention network for sequential recommendation. In Proceedings of the Twenty-Eighth International Joint Conference on Artificial Inte...

    2019

  84. [92]

    mgte: Generalized long-context text representa- tion and reranking models for multilingual text retrieval

    Xin Zhang, Yanzhao Zhang, Dingkun Long, Wen Xie, Ziqi Dai, Jialong Tang, Huan Lin, Baosong Yang, Pengjun Xie, Fei Huang, et al. mgte: Generalized long-context text representa- tion and reranking models for multilingual text retrieval. arXiv preprint arXiv:2407.19669, 2024

    2024 arXiv

  85. [93]

    Collm: Integrating collaborative embeddings into large language models for recommendation

    Yang Zhang, Fuli Feng, Jizhi Zhang, Keqin Bao, Qifan Wang, and Xiangnan He. Collm: Integrating collaborative embeddings into large language models for recommendation. CoRR, abs/2310.19488, 2023

    2023 arXiv

  86. [94]

    Adapting large language models by integrating collaborative semantics for recommendation

    Bowen Zheng, Yupeng Hou, Hongyu Lu, Yu Chen, Wayne Xin Zhao, Ming Chen, and Ji-Rong Wen. Adapting large language models by integrating collaborative semantics for recommendation. In 2024 IEEE 40th International Conference on Data Engineering (ICDE), 2024. 15

    2024

  87. [95]

    Llamafactory: Unified efficient fine-tuning of 100+ language models

    Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. Llamafactory: Unified efficient fine-tuning of 100+ language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System ...

    2024

  88. [96]

    A survey of recommender systems with multi-objective optimization

    Yong Zheng and David (Xuejun) Wang. A survey of recommender systems with multi-objective optimization. Neurocomputing, 2022

    2022

  89. [97]

    Deep interest network for click-through rate prediction

    Guorui Zhou, Xiaoqiang Zhu, Chenru Song, Ying Fan, Han Zhu, Xiao Ma, Yanghui Yan, Junqi Jin, Han Li, and Kun Gai. Deep interest network for click-through rate prediction. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages...

    2018

  90. [98]

    S3-rec: Self-supervised learning for sequential recommendation with mutual information maximization

    Kun Zhou, Hui Wang, Wayne Xin Zhao, Yutao Zhu, Sirui Wang, Fuzheng Zhang, Zhongyuan Wang, and Ji-Rong Wen. S3-rec: Self-supervised learning for sequential recommendation with mutual information maximization. In Proceedings of the 29th ACM international conference on informatio...

    1902

  91. [99]

    Joint optimization of tree-based index and deep model for recommender systems

    Han Zhu, Daqing Chang, Ziru Xu, Pengye Zhang, Xiang Li, Jie He, Han Li, Jian Xu, and Kun Gai. Joint optimization of tree-based index and deep model for recommender systems. Advances in Neural Information Processing Systems, 32, 2019

    2019

  92. [100]

    Learning tree-based deep model for recommender systems

    Han Zhu, Xiang Li, Pengye Zhang, Guozheng Li, Jie He, Han Li, and Kun Gai. Learning tree-based deep model for recommender systems. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 1079–1088, 2018

    2018

  93. [101]

    Cost: Contrastive quantization based semantic tokenization for generative recommendation

    Jieming Zhu, Mengqun Jin, Qijiong Liu, Zexuan Qiu, Zhenhua Dong, and Xiu Li. Cost: Contrastive quantization based semantic tokenization for generative recommendation. In RecSys, pages 969–974, 2024

    2024

  94. [102]

    Learning optimal tree models under beam search

    Jingwei Zhuo, Ziru Xu, Wei Dai, Han Zhu, Han Li, Jian Xu, and Kun Gai. Learning optimal tree models under beam search. In International Conference on Machine Learning, pages 11650–11659. PMLR, 2020. 16 A Appendix A.1 Limitations Computational Cost. In our efforts to deploy LLM...

    2020

  95. [103]

    By changing the query to a certain context, we can inject external world knowledge into URM, allowing the results to get ready for potential upcoming events

  96. [104]

    # !$%&'# Figure 10: Visualization of different item representations (case 1). Main Item Fusion !!

    This also supports the combination of URM and Chain-of-Thought (CoT) technologies [77]. Specifically, the LLM generates intermediate results in the form of text through reasoning, and then injects this text as context into URM, thereby producing the final set. Query: Christmas...

  97. [105]

    Guidelines: • The answer NA means that the abstract and introduction do not include the claims made in the paper

    Claims Question: Do the main claims made in the abstract and introduction accurately reflect the paper’s contributions and scope? Answer: [Yes] Justification: The abstract and introduction clearly state our contributions – introducing the Universal Retrieval Model (URM), enhan...

  98. [106]

    Limitations

    Limitations Question: Does the paper discuss the limitations of the work performed by the authors? Answer: [Yes] Justification: Due to space constraints, we have placed the limitations section in the Appendix A.1. Guidelines: • The answer NA means that the paper has no limitat...

  99. [107]

    Guidelines: • The answer NA means that the paper does not include theoretical results

    Theory assumptions and proofs 28 Question: For each theoretical result, does the paper provide the full set of assumptions and a complete (and correct) proof? Answer: [Yes] Justification: We provide detailed proof in Appendix A.3 for the assumption in Algorithm 1. Guidelines: ...

  100. [108]

    In Section 5.1,5.2, Appendix B.1,B.2,B.3, we provide all the details to reproduce our experiment

    Experimental result reproducibility Question: Does the paper fully disclose all the information needed to reproduce the main ex- perimental results of the paper to the extent that it affects the main claims and/or conclusions of the paper (regardless of whether the code and da...

  101. [109]

    Due to data security concerns, our code for the industrial online and offline experiments cannot be made public

    Open access to data and code Question: Does the paper provide open access to the data and code, with sufficient instruc- tions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [No] Justification: Upon acceptance, we will rel...

  102. [110]

    The full details can be found in Appendix B.2 and B.3

    Experimental setting/details Question: Does the paper specify all the training and test details (e.g., data splits, hyper- parameters, how they were chosen, type of optimizer, etc.) necessary to understand the results? Answer: [Yes] Justification: The experimental setting is p...

  103. [111]

    Nonetheless, our method consistently outperforms baseline methods across multiple tasks using both public and industrial datasets, showing marked improvements in online performance

    Experiment statistical significance Question: Does the paper report error bars suitably and correctly defined or other appropriate information about the statistical significance of the experiments? Answer: [No] Justification: Given the high computational cost, we did not inclu...

  104. [112]

    Guidelines: • The answer NA means that the paper does not include experiments

    Experiments compute resources Question: For each experiment, does the paper provide sufficient information on the com- puter resources (type of compute workers, memory, time of execution) needed to reproduce the experiments? Answer: [Yes] Justification: The computer resources ...

  105. [113]

    Guidelines: • The answer NA means that the authors have not reviewed the NeurIPS Code of Ethics

    Code of ethics Question: Does the research conducted in the paper conform, in every respect, with the NeurIPS Code of Ethics https://neurips.cc/public/EthicsGuidelines? Answer: [Yes] Justification: Our manuscript adheres to the NeurIPS Code of Ethics by ensuring equitable trea...

  106. [114]

    There may be concerns regarding the use of sensitive user or item information; however, we address this by omitting such details and utilizing non-sensitive attributes in practice

    Broader impacts Question: Does the paper discuss both potential positive societal impacts and negative societal impacts of the work performed? Answer: [Yes] 31 Justification: We discuss potential positive societal impacts in Appendix A.2, and we foresee no negative societal im...

  107. [115]

    Guidelines: • The answer NA means that the paper poses no such risks

    Safeguards Question: Does the paper describe safeguards that have been put in place for responsible release of data or models that have a high risk for misuse (e.g., pretrained language models, image generators, or scraped datasets)? Answer: [NA] Justification: Our paper poses...

  108. [116]

    Guidelines: • The answer NA means that the paper does not use existing assets

    Licenses for existing assets Question: Are the creators or original owners of assets (e.g., code, data, models), used in the paper, properly credited and are the license and terms of use explicitly mentioned and properly respected? Answer: [Yes] Justification: We have cited al...

  109. [117]

    Guidelines: • The answer NA means that the paper does not release new assets

    New assets Question: Are new assets introduced in the paper well documented and is the documentation provided alongside the assets? Answer: [NA] Justification: Our paper does not release new assets. Guidelines: • The answer NA means that the paper does not release new assets. ...

  110. [118]

    Guidelines: • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects

    Crowdsourcing and research with human subjects Question: For crowdsourcing experiments and research with human subjects, does the paper include the full text of instructions given to participants and screenshots, if applicable, as well as details about compensation (if any)? A...

  111. [119]

    Guidelines: 33 • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects

    Institutional review board (IRB) approvals or equivalent for research with human subjects Question: Does the paper describe potential risks incurred by study participants, whether such risks were disclosed to the subjects, and whether Institutional Review Board (IRB) approvals...

  112. [120]

    Answer: [Yes] Justification: Section 3.1 provides the inputs and outputs of the LLMs and a detailed explanation of the modifications we made to the existing LLMs

    Declaration of LLM usage Question: Does the paper describe the usage of LLMs if it is an important, original, or non-standard component of the core methods in this research? Note that if the LLM is used only for writing, editing, or formatting purposes and does not impact the ...

    2025

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.