Pith. sign in

REVIEW 4 major objections 5 minor 68 references

MRP-LLM: Multitask Reflective Large Language Models for Privacy-Preserving Next POI Recommendation

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Distilling user preferences by multitask prompting and self-reflection beats zero-shot prompting for LLM next-POI recommendation; adding differential privacy to all uploaded data costs 1.3% in accuracy and improves MRR by 0.8%.

desk verdict The non-private MR-LLM pipeline is a plausible empirical contribution, but the privacy-preserving variant as described cannot produce its reported utility: OUE at ε=0.1 destroys the category signal before it reaches the LLM. read the letter →

arxiv 2412.07796 v1 pith:UQNIK3JK submitted 2024-12-03 cs.IR cs.AI

Ziqing Wu , Zhu Sun , Dongxia Wang , Lu Zhang , Jie Zhang , Yew Soon Ong This is my paper · ORCID
classification cs.IRcs.AI
keywords nextPOIrecommendationlargelanguagemodelsprivacypreservationdifferentialin-contextlearningself-reflectioncollaborativesignalslocation-basedsocialnetworks
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

Most LLM-based next-POI recommenders simply hand a user's check-in history to the model and ask for the next place, which wastes the model's reasoning and exposes sensitive location data. This paper argues that a better route is to first prompt the LLM to distill a user's category, region, and distance preferences through multiple small prediction tasks, then to let the LLM correct its own preference statements against real check-ins, and finally to feed both the user's and similar users' summarized preferences into the final recommendation prompt. Across three real-world datasets, this multitask reflective pipeline lifts accuracy by 8.4% and mean reciprocal rank by 7.0% on average over the strongest zero-shot LLM baseline. The authors further show that when every uploaded data type is perturbed for differential privacy, the protected version still matches that baseline, losing only 1.3% in accuracy and gaining 0.8% in mean reciprocal rank. The upshot is that LLM reasoning for next-POI recommendation can survive a privacy-preserving data pipeline.

What carries the argument

The load-bearing mechanism is the fine-grained preference knowledge base built by Multitask Reflective Preference Extraction, together with the neighbor preference retrieval that injects collaborative signals into the final prompt. Each user's category, region, and distance sequences are probed with three subtasks, and the LLM's preferences are corrected against ground-truth check-ins from sampled recent and contextually relevant segments, producing five preference types that are stored per user in the knowledge base. At inference, the module selects the closest geographical, semantic, and social neighbors, summarizes their stored preferences, and appends them to the recommendation prompt. The privacy side is carried by four perturbation schemes applied before data leaves the device: optimized unary encoding for one-hot category, region, and distance records; Laplace noise for the distributions used in neighbor retrieval; random flipping for social links; and (rho,h)-privacy geo-fuzzification, which replaces each check-in POI with a nearby POI inside a radius chosen from check-in density.

What would settle it

Run MRP-LLM on the three datasets with the privacy budget epsilon set so low that optimized unary encoding outputs are nearly random, Laplace noise swamps the distributions, social links are uniformly flipped, and POI replacements are drawn from the widest allowed circle, then compare ACC@1 and MRR against the random-candidate baseline. If accuracy does not collapse toward that baseline, the perturbation is not actually removing the preference signal that drives the results, and the claimed privacy-utility trade-off would be an artifact.

Watch

Extended reading notes

Core claim

The central discovery is that a next-POI recommender built on a large language model performs substantially better when the model is made to reason about the user's fine-grained preferences before choosing a POI. The paper decomposes preference into categorical transition, categorical temporal, regional transition, regional temporal, and distance temporal preferences, extracts them with multitask prompts, refines them through self-reflection on recent and contextually similar check-in segments, and stores the results in a knowledge base. At recommendation time, it retrieves geographical, semantic, and social neighbors, summarizes their preferences, and injects both the user's own and the neighbors' preferences as hints for predicting the next category, region, distance, and finally the POI itself. The authors report that this design, without privacy protection, outperforms the strongest zero-shot LLM baseline LLMMove by an average of 8.4% in ACC and 7.0% in MRR on the Singapore, New York, and Phoenix datasets, while the privacy-preserving version, which perturbs all uploaded data with optimized unary encoding, Laplace noise, random link flipping, and geo-fuzzification, remains comparable to LLMMove with a 1.3% ACC drop and a 0.8% MRR gain.

Load-bearing premise

The privacy-preserving variant's utility claim depends on the assumption that the noise added for privacy still leaves enough genuine preference signal for the language model to extract sensible preferences and for neighbor lookup to find users who are truly similar.

Editorial extensions

If this is right

  • If the reported gains hold, LLM-based next-POI recommenders should move away from raw check-in prompts toward a distill-then-recommend structure, since the multitask reflective pipeline yields an average 8.4% ACC and 7.0% MRR lift over the strongest zero-shot LLM baseline without privacy protection.
  • Full differential privacy on all uploaded user data is affordable for this task: the protected model loses only 1.3% in ACC and gains 0.8% in MRR relative to the same zero-shot baseline, so users do not have to choose sharply between privacy and recommendation quality.
  • Collaborative signals can be expressed as summarized natural-language preference statements from similar users and injected into the prompt, meaning that in-context learning can exploit other users' behavior without exposing their raw histories.
  • Across the ablations, POI-sequence perturbation hurts accuracy most while distribution perturbation hurts least, so future privacy-utility engineering should focus on protecting exact POI coordinates more efficiently.
  • The privacy-utility trade-off is monotone in the privacy budget: higher epsilon (less noise) improves accuracy, giving deployments a tunable knob based on how much they trust the platform.

Reading between the lines

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

  • Beyond the paper, the same distill-retrieve-recommend structure could transfer to other sequential recommendation domains, such as next video or next purchase, wherever user choice decomposes into a few interpretable aspects and neighbor preferences can be summarized as text hints.
  • Beyond the paper, the neighbor-selection step could be attacked by an adversary who knows the noise mechanism and privacy budget and attempts to reconstruct true check-in distributions or the social graph from the published perturbed versions; the paper does not evaluate this stronger adversarial setting.
  • Beyond the paper, since the ablation shows POI-sequence perturbation is the most costly privacy operation, allocating privacy budget asymmetrically, with more budget for POI sequences and less for distributions, may yield a better privacy-utility frontier than the uniform epsilon used here.
  • Beyond the paper, because the pipeline is prompt-based, running it with an openly available language model as a drop-in replacement would clarify how much of the reported lift depends on the specific proprietary model's instruction-following ability.
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 MRP-LLM, an LLM-based next-POI recommendation framework with three main components: a multitask reflective preference extraction module that distills fine-grained categorical, regional, and distance preferences into a knowledge base; a neighbor preference retrieval module that injects collaborative signals; and a multitask next-POI recommendation module that uses the extracted preferences for prediction. A privacy transmission module is designed to protect check-in sequences, distributions, social links, and POI records using OUE, Laplace noise, random flipping, and geo-fuzzification. The authors report experiments on Foursquare data from Singapore, New York, and Phoenix, claiming that the non-private variant MR-LLM outperforms the best LLM baseline LLMMove by 8.4% ACC and 7.0% MRR, and that the privacy-preserving variant MRP-LLM remains comparable to LLMMove.

Significance. If the results hold, the paper would make two useful contributions to LLM-based POI recommendation: a structured, reflective preference-extraction pipeline that improves over zero-shot LLM prompting, and an attempt to incorporate local differential privacy into the data upload process. The paper is transparent about prompts, hyperparameter search spaces, and dataset statistics, and it evaluates on three real-world datasets. However, the privacy-preserving mechanism as described appears internally inconsistent, and the experimental reporting lacks the statistical detail needed to support the small performance differences between MRP-LLM and the baselines. The non-private MR-LLM result is interesting, but the headline privacy-utility claim is not currently established.

major comments (4)
  1. [§III-F.1, Eq. (3), Algorithm 2] The OUE perturbation is applied independently to each one-hot category/region/distance record, and the perturbed sequences C'_u, R'_u, and D'_u are passed directly to PreferenceExt (Algorithm 2 lines 3-6). At the reported optimal ε=0.1 and with d=398 categories for SIN, P(0→1)=1/(e^0.1+1)≈0.475, so a single perturbed record contains, in expectation, about 0.5 + 397×0.475 ≈ 189 ones. No decoding, aggregation, or thresholding step is specified before these vectors are fed to the LLM in Prompt 2. The true category is retained with probability only 0.5 and is buried among roughly 188 false positives, so with the short segments used in the paper (n=5, m=1-2) the categorical transition and temporal preferences cannot be recovered. Consequently, the MRP-LLM results in Table IV rows 10 and 14 cannot be produced by the privacy mechanism as described. The authors must specify a statistically valid decoding/inference procedure, or replace the perturbation scheme and rerun the privacy-preserving experiments.
  2. [Table III and Figure 7] The reported optimal privacy budget is ε=0.1, the smallest value in the searched range [0.1, 0.9], while Figure 7 shows accuracy increasing monotonically with ε on all three datasets. This is inconsistent with the statement in Section IV-D that the best hyperparameters were selected empirically. If ε=0.1 was chosen to maximize privacy rather than utility, the selection criterion must be stated and the sensitivity analysis should report results across the full range; otherwise the privacy-utility trade-off, and the specific operating point used in Table IV, are not characterized as claimed.
  3. [§IV-B, §IV-D, Table IV] The paper reports 10-run averages without standard deviations, confidence intervals, or significance tests. Many differences between MRP-LLM and LLMMove are small and inconsistent across datasets (e.g., NY row 14: -8.1% ACC@1, -6.5% ACC@5, -5.1% ACC@10, -6.3% MRR; PHO ACC@10 shows a 0% difference). The claims of 'comparable performance' and an average 0.8% MRR lift are therefore not statistically supported. In addition, the 100-POI candidate sampling protocol should be made reproducible by fixing seeds or releasing the candidate sets, and the paper should state whether the same candidate sets are used for all methods.
  4. [§III-F] The paper invokes differential privacy for four different mechanisms (OUE, Laplace, random flipping, and geo-fuzzification) but provides no composition analysis or formal privacy guarantee for the combined pipeline. In particular, Algorithm 1 refers to the '⟨ϱ, h⟩-privacy' technique without defining the privacy notion or stating how the budget ε in Algorithm 1 composes with the budgets used in Equations (3)-(5). Since privacy preservation is a central claimed contribution, the privacy analysis must be made rigorous before the privacy-preserving results can be evaluated.
minor comments (5)
  1. [Prompt 4] Prompt 4 asks the model to 'Generate the updated categorical transition preference' twice; the second question should presumably ask for the updated categorical temporal preference.
  2. [§IV-F and case study] There are several typos: 'Prmopt 1' should be 'Prompt 1', 'categroy' should be 'category', and 'LLMob' in Section II-B should be 'LLMMob' for consistency.
  3. [Eq. (5)] The random flipping mechanism in Equation (5) uses the condition p/q ≤ exp(ε) but does not specify the relationship between p and q or how the resulting mechanism satisfies differential privacy; please clarify the parameterization.
  4. [Algorithm 1] In Algorithm 1, line 6 'c′ ← random flip(ϵ)' is undefined; if this is a randomized response on the category attribute, the text should say so and specify the flip probability, since this affects both the utility and privacy of the POI replacement.
  5. [Table IV] The NY block of Table IV repeats the label 'Conventional Methods' above the LLM-based baseline rows 7-8; this is a copy-paste error that should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are held-out measurements against external baselines, and the preference KB is produced by LLM prompting rather than fitted to the target metric.

full rationale

I walked the paper's derivation chain. MR-LLM and MRP-LLM are evaluated by ACC@K and MRR on the last check-in of a held-out validation/test split (Section IV-A/B), using 100-POI candidate sets containing the ground truth; the same protocol is applied to all baselines. The preference KB is generated by feeding perturbed category/region/distance sequences through Prompts 2-4 and self-reflection (Section III-C, Algorithm 2); it is not constructed by optimizing the reported ACC/MRR, so the later recommendation prompt does not read back a fitted target value. Neighbor retrieval uses KL divergence over distributions and social links (Eqs. 1-2 and 5), independent of the target metric. Equation 3 (OUE), Equation 4 (Laplace), and Algorithm 1 (geo-fuzzing) are input perturbations applied before prompting; no reported number is defined in terms of a parameter fitted to that same number. Hyperparameters (m, n, rho, epsilon) are selected on validation (Table III), which is standard practice and not circular. The only self-citations (e.g., [8], [9], [11], [61]) are contextual related-work and factor-identification references, not load-bearing uniqueness claims. The OUE-versus-utility concern raised by the skeptic is an internal plausibility/correctness issue rather than a circularity: even if the perturbation destroys signal, that would invalidate the utility claim empirically, not make the claim true by definition. Therefore no circular step exists.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical or mathematical entities are introduced. The preference KB is an internal data structure, not an independent postulated entity. The free parameters are standard hyperparameters and hand-set constraints, and the central claim rests on domain assumptions about LLM preference extraction and perturbation preserving signal.

free parameters (5)
  • m (number of reflective segments per user) = SIN=1, NY=1, PHO=2
    Tuned on validation data, Table III and Figure 4.
  • n (segment length) = SIN=5, NY=3, PHO=5
    Tuned on validation data, Table III and Figure 5.
  • rho (preference extraction participation rate) = 1.0 on all datasets
    Tuned on validation data, Table III and Figure 6.
  • epsilon (privacy budget) = 0.1 on all datasets
    Tuned on validation data, Table III and Figure 7.
  • geo-fuzzification radius range = 10 km to 30 km
    Hand-set in Section III-F to keep enough POIs inside the replacement circle; not tuned.
assumptions (4)
  • domain assumption LLM prompting on short check-in segments can extract stable, useful categorical, regional, and distance preferences.
    Used throughout Section III-C; if false, the preference KB would not support the reported accuracy gains.
  • domain assumption KL divergence between regional or categorical check-in distributions identifies users whose preferences are useful collaborative signals.
    Section III-D, Eqs. 1-2; no comparison with alternative similarity measures.
  • domain assumption The four perturbation mechanisms (OUE, Laplace, random flipping, geo-fuzzification) each provide differential privacy as stated, and their composition is acceptable.
    Section III-F invokes standard mechanisms but provides no end-to-end privacy budget accounting.
  • domain assumption The Foursquare datasets preprocessed with 5-core filtering and day-based splitting form a fair evaluation benchmark.
    Section IV-A; follows prior work but no distributional analysis is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MRP-LLM: Multitask Reflective Large Language Models for Privacy-Preserving Next POI Recommendation." pith.science (2026). https://pith.science/paper/UQNIK3JK

@misc{pith2026241207796,
  author       = {Pith},
  title        = {Pith review of: MRP-LLM: Multitask Reflective Large Language Models for Privacy-Preserving Next POI Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UQNIK3JK}},
  note         = {Machine review of arXiv:2412.07796}
}
read the original abstract

Large language models (LLMs) have shown promising potential for next Point-of-Interest (POI) recommendation. However, existing methods only perform direct zero-shot prompting, leading to ineffective extraction of user preferences, insufficient injection of collaborative signals, and a lack of user privacy protection. As such, we propose a novel Multitask Reflective Large Language Model for Privacy-preserving Next POI Recommendation (MRP-LLM), aiming to exploit LLMs for better next POI recommendation while preserving user privacy. Specifically, the Multitask Reflective Preference Extraction Module first utilizes LLMs to distill each user's fine-grained (i.e., categorical, temporal, and spatial) preferences into a knowledge base (KB). The Neighbor Preference Retrieval Module retrieves and summarizes the preferences of similar users from the KB to obtain collaborative signals. Subsequently, aggregating the user's preferences with those of similar users, the Multitask Next POI Recommendation Module generates the next POI recommendations via multitask prompting. Meanwhile, during data collection, a Privacy Transmission Module is specifically devised to preserve sensitive POI data. Extensive experiments on three real-world datasets demonstrate the efficacy of our proposed MRP-LLM in providing more accurate next POI recommendations with user privacy preserved.

Figures

Figures reproduced from arXiv: 2412.07796 by the authors.

Figure 1
Figure 1. Our proposed method shown in (b), which makes recommendations [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The architecture of MRP-LLM. Taking u1 as the recommendation query user. ask it to extract the user’s categorical transition and temporal preferences in Prompt 2. Using the same approach, we then probe the user’s regional transition and temporal preferences with Rcur u and her distance temporal preference based on Dcur u , step by step. Prompt 1: Task Instruction Your task is to recommend a user’s next point-of￾inte… view at source ↗
Figure 3
Figure 3. Results of ablation study. 7) LLMMob: [10] performing zero-shot recommendation via LLMs based on users’ history and current sequences; 8) LLMMove: [12] performing zero-shot recommendation using LLMs based on users’ check-in sequences, dis￾tance, and transition patterns. D. Implementation Details We empirically find all the best hyper-parameters for all methods on the three datasets. For each evaluation, we ran￾domly… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Results of parameter sensitivity analysis for segment number [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Results of parameter sensitivity analysis for segment length [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Results of parameter sensitivity analysis for preference extraction participation rate [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Results of parameter sensitivity analysis for privacy budget [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 60 canonical work pages

  1. [1]

    Point-of-interest recommender systems based on location-based social networks: a survey from an experimental perspec- tive,

    P. S ´anchez et al. , “Point-of-interest recommender systems based on location-based social networks: a survey from an experimental perspec- tive,” CSUR, vol. 54, pp. 1–37, 2022

    work page 2022

  2. [2]

    Factorizing personalized markov chains for next-basket recommendation,

    S. Rendle et al., “Factorizing personalized markov chains for next-basket recommendation,” in TheWebConf, 2010, pp. 811–820

    2010

  3. [3]

    Predicting the next location: a recurrent model with spatial and temporal contexts,

    Q. Liu et al., “Predicting the next location: a recurrent model with spatial and temporal contexts,” in AAAI, 2016, pp. 194–200

    work page 2016

  4. [4]

    Stan: spatio-temporal attention network for next location recommendation,

    Y . Luo et al., “Stan: spatio-temporal attention network for next location recommendation,” in WWW, 2021, pp. 2177–2185

    work page 2021

  5. [5]

    Hierarchical multi-task graph recurrent network for next poi recommendation,

    N. Lim et al., “Hierarchical multi-task graph recurrent network for next poi recommendation,” in SIGIR, 2022, pp. 1133–1143

    work page 2022

  6. [6]

    Language models represent space and time,

    W. Gurnee et al., “Language models represent space and time,” in ICLR, 2024

    work page 2024

  7. [7]

    Drdt: dynamic reflection with divergent thinking for llm-based sequential recommendation,

    Y . Wang et al. , “Drdt: dynamic reflection with divergent thinking for llm-based sequential recommendation,” arXiv, 2023

    work page 2023

  8. [8]

    Adaptive in-context learning with large language models for bundle generation,

    Z. Sun et al., “Adaptive in-context learning with large language models for bundle generation,” in SIGIR, 2024, pp. 966–976

    work page 2024

Show all 68 references
  1. [9]

    Large language models for intent-driven session recommenda- tions,

    ——, “Large language models for intent-driven session recommenda- tions,” in SIGIR, 2024, pp. 324–334

    2024

  2. [10]

    Where would i go next? large language models as human mobility predictors,

    X. Wang et al. , “Where would i go next? large language models as human mobility predictors,” arXiv, 2023

    2023

  3. [11]

    An interactive multi-task learning framework for next poi recommendation with uncertain check-ins,

    L. Zhang et al., “An interactive multi-task learning framework for next poi recommendation with uncertain check-ins,” in IJCAI, 2020, pp. 3551–3557

    2020

  4. [12]

    Where to move next: zero-shot generalization of llms for next poi recommendation,

    S. Feng et al. , “Where to move next: zero-shot generalization of llms for next poi recommendation,” in CAI. IEEE, 2024, pp. 1530–1535

    2024

  5. [13]

    Willingness to share data: contextual deter- minants of consumers’ decisions to share private data with companies,

    K. A. Ackermann et al. , “Willingness to share data: contextual deter- minants of consumers’ decisions to share private data with companies,” Journal of Consumer Behaviour , vol. 21, pp. 375–386, 2022

    2022

  6. [14]

    Where you like to go next: successive point-of-interest recommendation,

    C. Cheng et al., “Where you like to go next: successive point-of-interest recommendation,” in IJCAI, 2013, pp. 2605––2611

    2013

  7. [15]

    Fundamentals of recurrent neural network (rnn) and long short-term memory (lstm) network,

    A. Sherstinsky, “Fundamentals of recurrent neural network (rnn) and long short-term memory (lstm) network,” Physica D: Nonlinear Phe- nomena, vol. 404, pp. 132–306, 2020

    2020

  8. [16]

    Where to go next: a spatio-temporal gated network for next poi recommendation,

    P. Zhao et al., “Where to go next: a spatio-temporal gated network for next poi recommendation,” TKDE, vol. 34, pp. 2512–2524, 2020

    2020

  9. [17]

    Long short-term memory,

    S. Hochreiter et al. , “Long short-term memory,” Neural Computation , vol. 9, pp. 1735–1780, 1997

    1997

  10. [18]

    Attention is all you need,

    A. Vaswani et al., “Attention is all you need,” NeurIPS, vol. 30, 2017

    2017

  11. [19]

    Deepmove: predicting human mobility with attentional recurrent networks,

    J. Feng et al. , “Deepmove: predicting human mobility with attentional recurrent networks,” in TheWebConf, 2018, pp. 1459–1468

    2018

  12. [20]

    An attention-based spatiotemporal lstm network for next poi recommendation,

    L. Huang et al. , “An attention-based spatiotemporal lstm network for next poi recommendation,” TSC, vol. 14, pp. 1585–1597, 2019

    2019

  13. [21]

    Next point-of-interest recommendation with inferring multi-step future preferences

    L. Zhang et al., “Next point-of-interest recommendation with inferring multi-step future preferences.” in IJCAI, 2022, pp. 3751–3757

    2022

  14. [22]

    Spatio-temporal hypergraph learning for next poi recom- mendation,

    X. Yan et al., “Spatio-temporal hypergraph learning for next poi recom- mendation,” in SIGIR, 2023, pp. 403–412

    2023

  15. [23]

    Disentangled contrastive hypergraph learning for next poi recommendation,

    Y . Lai et al., “Disentangled contrastive hypergraph learning for next poi recommendation,” in SIGIR, 2024, pp. 1452–1462

    2024

  16. [24]

    Is chatgpt a good recommender? a preliminary study,

    J. Liu et al. , “Is chatgpt a good recommender? a preliminary study,” arXiv, 2023

    2023

  17. [25]

    Zero-shot next-item recommendation using large pretrained language models,

    L. Wang et al. , “Zero-shot next-item recommendation using large pretrained language models,” arXiv, 2023

    2023

  18. [26]

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

    J. Wei et al. , “Chain-of-thought prompting elicits reasoning in large language models,” NeurIPS, vol. 35, pp. 24 824–24 837, 2022

    2022

  19. [27]

    Is chatgpt good at search? investigating large language models as re-ranking agents,

    W. Sun et al., “Is chatgpt good at search? investigating large language models as re-ranking agents,” in EMNLP, 2023, pp. 14 918–14 937

    2023

  20. [28]

    Towards open-world recommendation with knowledge augmentation from large language models,

    Y . Xi et al. , “Towards open-world recommendation with knowledge augmentation from large language models,” in Recsys, 2024, pp. 12– 22

    2024

  21. [29]

    Llmrec: large language models with graph augmentation for recommendation,

    W. Wei et al., “Llmrec: large language models with graph augmentation for recommendation,” in WSDM, 2024, pp. 806–815

    2024

  22. [30]

    Large language model with graph convolution for recommendation,

    Y . Du et al. , “Large language model with graph convolution for recommendation,” arXiv, 2024

    2024

  23. [31]

    Generative sequential recommendation with gptrec,

    A. V . Petrov et al., “Generative sequential recommendation with gptrec,” arXiv preprint arXiv:2306.11114 , 2023

    2023 arXiv

  24. [32]

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

    S. Geng et al., “Recommendation as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5),” in Recsys, 2022, pp. 299–315

    2022

  25. [33]

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

    K. Bao et al. , “Tallrec: an effective and efficient tuning framework to align large language model with recommendation,” in Recsys, 2023, pp. 1007–1014

    2023

  26. [34]

    Once: boosting content-based recommendation with both open-and closed-source large language models,

    Q. Liu et al., “Once: boosting content-based recommendation with both open-and closed-source large language models,” in WSDM, 2024, pp. 452–461

    2024

  27. [35]

    Chat-rec: towards interactive and explainable llms- augmented recommender system,

    Y . Gao et al. , “Chat-rec: towards interactive and explainable llms- augmented recommender system,” arXiv, 2023

    2023

  28. [36]

    Large language models are zero-shot rankers for recommender systems,

    Y . Hou et al. , “Large language models are zero-shot rankers for recommender systems,” in ECIR, 2024, pp. 364–381

    2024

  29. [37]

    Re2llm: reflective reinforcement large language model for session-based recommendation,

    Z. Wang et al., “Re2llm: reflective reinforcement large language model for session-based recommendation,” arXiv, 2024

    2024

  30. [38]

    Large language models for next point-of-interest recom- mendation,

    P. Li et al. , “Large language models for next point-of-interest recom- mendation,” in SIGIR, 2024, pp. 1463–1472

    2024

  31. [39]

    Lightweight privacy-preserving distributed recommender system using tag-based multikey fully homomorphic data encapsulation,

    J. Zhou et al., “Lightweight privacy-preserving distributed recommender system using tag-based multikey fully homomorphic data encapsulation,” IEEE Transactions on Dependable and Secure Computing, vol. 20, no. 6, pp. 5230–5246, 2023

    2023

  32. [40]

    Towards privacy-preserving category-aware poi recom- mendation over encrypted lbsn data,

    L. Sun et al. , “Towards privacy-preserving category-aware poi recom- mendation over encrypted lbsn data,” Inf. Sci., vol. 662, p. 120253, 2024

    2024

  33. [41]

    The algorithmic foundations of differential privacy,

    C. Dwork et al. , “The algorithmic foundations of differential privacy,” Found. Trends Theor. Comput. Sci., vol. 9, pp. 211–407, 2014

    2014

  34. [42]

    Model-agnostic decentralized collaborative learning for on-device poi recommendation,

    J. Long et al., “Model-agnostic decentralized collaborative learning for on-device poi recommendation,” in SIGIR, 2023, pp. 423–432

    2023

  35. [43]

    Fedpoirec: privacy-preserving federated poi recom- mendation with social influence,

    V . Perifanis et al., “Fedpoirec: privacy-preserving federated poi recom- mendation with social influence,” Inf. Sci., vol. 623, pp. 767–790, 2023

    2023

  36. [44]

    Ranking-based federated poi recommendation with geographic effect,

    Q. Dong et al. , “Ranking-based federated poi recommendation with geographic effect,” in IJCNN, 2022, pp. 1–8

    2022

  37. [45]

    Decentralized collaborative learning framework for next poi recommendation,

    J. Long et al., “Decentralized collaborative learning framework for next poi recommendation,” TIST, vol. 41, no. 3, pp. 1–25, 2023

    2023

  38. [46]

    Pad: privacy-area aware, dummy-based location privacy in mobile services,

    H. Lu et al. , “Pad: privacy-area aware, dummy-based location privacy in mobile services,” in MobiDE, 2008, pp. 16–23

    2008

  39. [47]

    Privacy-preserving point-of-interest recommendation based on geographical and social influence,

    Y . Huo et al. , “Privacy-preserving point-of-interest recommendation based on geographical and social influence,” Inf. Sci., vol. 543, pp. 202– 218, 2021

    2021

  40. [48]

    Cybercrime and privacy threats of large language models,

    N. Kshetri, “Cybercrime and privacy threats of large language models,” IT Professional, vol. 25, no. 3, pp. 9–13, 2023

    2023

  41. [49]

    Text revealer: private text reconstruction via model in- version attacks against transformers,

    R. Zhang et al., “Text revealer: private text reconstruction via model in- version attacks against transformers,” arXiv preprint arXiv:2209.10505, 2022

    2022 arXiv

  42. [50]

    Quantifying privacy risks of masked language models using membership inference attacks,

    F. Mireshghallah et al., “Quantifying privacy risks of masked language models using membership inference attacks,” in EMNLP, 2022, pp. 8332–8347

    2022

  43. [51]

    Ew-tune: A framework for privately fine-tuning large language models with differential privacy,

    R. Behnia et al., “Ew-tune: A framework for privately fine-tuning large language models with differential privacy,” in ICDMW. IEEE, 2022, pp. 560–566

    2022

  44. [52]

    Just fine-tune twice: selective differential privacy for large language models,

    W. Shi et al., “Just fine-tune twice: selective differential privacy for large language models,” in EMNLP, 2022, pp. 6327–6340

    2022

  45. [53]

    Fedbpt: efficient federated black-box prompt tuning for large language models,

    J. Sun et al. , “Fedbpt: efficient federated black-box prompt tuning for large language models,” in ICML, 2023

    2023

  46. [54]

    Propile: probing privacy leakage in large language models,

    S. Kim et al. , “Propile: probing privacy leakage in large language models,” in NeurIPS, 2024

    2024

  47. [55]

    Bumblebee: secure two-party inference framework for large transformers,

    W.-j. Lu et al., “Bumblebee: secure two-party inference framework for large transformers,” Cryptology ePrint Archive , 2023

    2023

  48. [56]

    Puma: secure inference of llama-7b in five minutes,

    Y . Dong et al. , “Puma: secure inference of llama-7b in five minutes,” arXiv preprint arXiv:2307.12533 , 2023

    2023

  49. [57]

    Split-and-denoise: protect large language model inference with local differential privacy,

    P. Mai et al., “Split-and-denoise: protect large language model inference with local differential privacy,” in ICML, 2024

    2024

  50. [58]

    Selective differential privacy for language modeling,

    W. Shi et al., “Selective differential privacy for language modeling,” in NAACL, 2022, pp. 2848–2859

    2022

  51. [59]

    Privacy-preserving prompt tuning for large language model services,

    Y . Li et al., “Privacy-preserving prompt tuning for large language model services,” arXiv, 2023

    2023

  52. [60]

    Privacy-preserving recommender systems with synthetic query generation using differentially private large language models,

    A. G. Carranza et al. , “Privacy-preserving recommender systems with synthetic query generation using differentially private large language models,” arXiv, 2023

    2023

  53. [61]

    A multi-channel next poi recommendation framework with multi-granularity check-in signals,

    Z. Sun et al. , “A multi-channel next poi recommendation framework with multi-granularity check-in signals,” TIST, vol. 42, no. 1, pp. 1–28, 2023

    2023

  54. [62]

    Self-refine: iterative refinement with self-feedback,

    A. Madaan et al., “Self-refine: iterative refinement with self-feedback,” NeurIPS, vol. 36, 2024

    2024

  55. [63]

    On information and sufficiency,

    S. Kullback et al., “On information and sufficiency,” Ann. Math. Stat. , vol. 22, no. 1, pp. 79–86, 1951

    1951

  56. [64]

    Locally differentially private protocols for frequency estimation,

    T. Wang et al. , “Locally differentially private protocols for frequency estimation,” in USENIX Security Symposium , 2017, pp. 729–745

    2017

  57. [65]

    Dplcf: differentially private local collaborative filtering,

    C. Gao et al., “Dplcf: differentially private local collaborative filtering,” in SIGIR, 2020, pp. 961–970

    2020

  58. [66]

    Participatory cultural mapping based on collective behavior data in location-based social networks,

    D. Yang et al. , “Participatory cultural mapping based on collective behavior data in location-based social networks,” TIST, vol. 7, no. 3, pp. 1–23, 2016

    2016

  59. [67]

    Next point-of-interest recommendation with inferring multi-step future preferences,

    L. Zhang et al., “Next point-of-interest recommendation with inferring multi-step future preferences,” in IJCAI, 2022, pp. 3751–3757

    2022

  60. [68]

    Bpr: bayesian personalized ranking from implicit feedback,

    S. Rendle et al. , “Bpr: bayesian personalized ranking from implicit feedback,” in UAI, 2012, pp. 452–461

    2012

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.