Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Active Large Language Model-based Knowledge Distillation for Session-based Recommendation

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read 500 hand-picked sessions can beat full LLM distillation

desk verdict New three-way active distillation idea for LLM-based SBR, but the theoretical guarantee doesn't match the implemented policy. read the letter →

arxiv 2502.15685 v1 pith:FOJ2YY2S submitted 2024-12-15 cs.IR cs.LG

classification cs.IRcs.LG
keywords session-basedrecommendationknowledgedistillationlargelanguagemodelsactivelearningmax-minoptimizationrankingstudent-teachersustainableAI
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

Session-based recommenders that distill knowledge from large language models normally ask the LLM to score every training session, which is expensive and noisy. The paper claims that a small student model can instead learn from a few hundred sessions picked by an active-learning rule that maximizes the minimal expected gain. The rule sorts sessions by three possible outcomes of an LLM prediction—effective, similar, or incorrect—and weights the sampling so that even the worst-case mix of these outcomes yields a guaranteed positive distillation gain. On two real-world datasets the method improves over prior knowledge-distillation baselines across three backbone recommenders while querying the LLM only about 500 times. If true, this makes LLM-based distillation practical for on-device session-based recommendation at a fraction of the API cost.

What carries the argument

The load-bearing object is the max-min equilibrium distribution $p^*$: each session's sampling probability is proportional to $1/(g^{ef}_s + g^{in}_s)$ or $1/(g^{si}_s + g^{in}_s)$ depending on its rank, with a cutoff $k^*$ that maximizes the closed-form value $\Gamma(k^*)$. Difficulty is measured by the negative consistency between the session embedding and item embeddings in a conventional teacher, and the gains for effective, similar, and incorrect predictions are derived from that difficulty. This machinery converts an intractable optimization over unknown instance labels into a solvable max-min problem, which is what carries the argument.

What would settle it

On a new dataset with ground-truth session labels, label each LLM prediction as effective, similar, or incorrect, compute $p^*$ from the true counts, and compare it against the grid-searched 1:5:4 policy; if the true-count policy does not outperform the grid-searched one, or if the grid-searched policy falls below random selection as the true mix diverges from 1:5:4, the central claim is weakened.

Watch

Extended reading notes

Core claim

The paper's central claim is that the active selection policy $p^*$ obtained by maximizing the minimal expected gain selects instances that are as informative as possible while avoiding incorrect or similar LLM predictions. The paper models distillation gain as a max-min game: the selector chooses a sampling distribution over sessions, and an adversary decides which sessions are effective, similar, or incorrect subject only to overall counts; Theorems 1--3 locate the equilibrium and give the closed-form value $\Gamma(k^*)$. It then shows empirically that this policy, implemented as ALKDRec with a GPT-4-turbo teacher and only 500--750 predicted sessions, beats all tested KD baselines on Hetrec2011-ML and Amazon-Games, and in several configurations outperforms a teacher recommender with ten times the latent dimension.

Load-bearing premise

The guarantee holds only if the total numbers of effective, similar, and incorrect LLM predictions are known in advance; the experiments do not measure these counts but fix a 1:5:4 ratio by grid search, so a dataset with a different mix could weaken the guarantee.

Editorial extensions

If this is right

  • Querying the LLM teacher for about 500 sessions instead of all sessions cuts the reported API cost from roughly 347 USD to under 9 USD and wall-clock time from about 30 hours to 44 minutes on Amazon-Games.
  • A student with latent dimension 10 can match or beat a teacher with latent dimension 100 after distillation, which brings LLM-level ranking quality to devices with small memory budgets.
  • Because the policy samples from the whole gain-ranked list rather than picking only the hardest or easiest sessions, it avoids both redundant and misleading training signals that degrade existing KD baselines.
  • The active-learning module is backbone-agnostic and improves FPMC, STAMP, and AttMix students, so the recipe can be composed with other session-based recommenders.
  • The max-min formulation needs only aggregate counts of the three prediction types, not per-instance labels, so the method does not have to know which LLM predictions are correct in advance.

Reading between the lines

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

  • An extension the paper leaves implicit: the 1:5:4 ratio is a hyperparameter, so a production system could estimate the three counts on a small probe set and recompute $p^*$, which would preserve the theoretical guarantee on new data.
  • The same three-outcome max-min decomposition should transfer to distilling from any expensive teacher in ranking tasks, such as vision-language models or ensembles, not only LLM recommenders.
  • A testable prediction of the theory is that the optimal subset size grows with the fraction of effective instances and shrinks as noise rises; the paper's single-dataset study of subset size (peak near 500) is consistent with but does not prove that trade-off.
  • The paper's stated future work—embedding-level distillation—suggests the active policy could also select sessions for hidden-state alignment; selecting for both ranking and embedding gains might close more of the gap to the LLM teacher.
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

3 major / 4 minor

Summary. The paper proposes ALKDRec, an active knowledge distillation method for session-based recommendation. A conventional teacher recommender is first trained on all sessions; its predictions are summarized by an LLM, which is then prompted to produce ranking lists for candidate items. To limit LLM inference cost, only a subset of sessions is selected for distillation: the selection policy is obtained by maximizing the minimum expected gain, where gains depend on an instance's difficulty and on whether the LLM prediction is effective, similar, or incorrect. The authors prove (Theorems 1-3) a max-min characterization of the optimal policy, and they report experiments on Hetrec2011-ML and Amazon-Games with FPMC, STAMP, and AttMix backbones, showing that ALKDRec improves over KD baselines in most configurations and is much cheaper than running the LLM on all sessions.

Significance. If the theoretical guarantee were intact, the paper would make a useful contribution: it extends robust active distillation from binary classification to ranking-oriented session-based recommendation with three outcome types, and it shows a practical path for a lightweight student to absorb LLM knowledge from a few hundred sessions. The empirical package is a genuine strength: real-world datasets, three backbones, statistical significance tests, efficiency and cost numbers, and ablations against Random, Easiest, Hardest, and RAD-BC. The main reservation is that the max-min guarantee is invoked under preconditions that the implementation does not verify, and the proof sketches contain enough index and notation errors that the claimed result cannot be checked as written.

major comments (3)
  1. [Active learning for LLM-based distillation, Eq. (3); Implementation Details] The max-min problem in Eq. (3) and Theorems 1-3 require the true totals kef, ksi, and kin of effective, similar, and incorrect instances. The text states that only these totals are assumed known, but the implementation does not measure them: 'We set the number of effective/similar/incorrect instances as 1:5:4 for all datasets by the grid search.' The resulting policy p* is therefore optimal for a different assignment set than the one containing the actual data, and Theorem 3 gives no lower bound on the realized expected gain when the true ratio differs from 1:5:4. Since this is the load-bearing justification for the claim that ALKDRec 'can select effective instances,' the authors should either measure these counts on validation data and report their stability, formulate the theorem over a family of possible counts, or explicitly reposition the max-min solution as a heuristic whose value is empirical.
  2. [Definition 1, Definition 2, and Proofs of Theorems 1-2] The proof sketches contain errors that prevent verification of the central claim. In Definition 1, the second branch of H_k sums from s=1 to k instead of from k_si+k_in+1 to k, in both the main text and Appendix B. In Theorem 1's proof, the minimization step says we should set c^ef_s=1 for s >= k* because p*_s=0, but p*_s=0 only for s > k*. In Definition 2, the denominators for hat c^ef and hat c^si appear swapped relative to the subsequent algebra in the proof of Theorem 2: the first branch of hat c^si is defined with g^ef_s+g^in_s, yet the proof substitutes (Γ(k*)+g^in_s)/(g^si_s+g^in_s) for that branch. These are not merely typographical: they obscure whether the lower-bound and upper-bound arguments are valid. The manuscript needs a complete, consistently labeled proof (or a correctly typeset appendix) for the max-min theorem before the theoretical contribution can be accepted.
  3. [Implementation Details; Figure 3; Ablation Study (RQ3)] The gain values used in practice are set by hand: g^ef_s = 1/[rank(d^f_s)]^mu with mu=10, g^si_s = g^in_s = g^ef_s/2, and the three counts are grid-searched to 1:5:4 on the same validation sets used to select hyperparameters. Because the theory optimizes the expected value with respect to these user-supplied gains, the theorem does not by itself establish that the selected subset maximizes actual knowledge-distillation utility; it only establishes robustness relative to the chosen gain model. The empirical comparison in RQ3 cannot fully separate the benefit of the max-min principle from the benefit of this validation-set tuning, since the Random, Easiest, Hardest, and RAD-BC baselines are all evaluated under the same hand-set gains. I would like to see a sensitivity analysis over the gain ratio and the count ratio (beyond Figure 3) and, ideally, a calibration check showing that the chosen ratios approximate the true counts on a held-out part of the data.
minor comments (4)
  1. [Before Algorithm 1] The sentence ending '...with k* shown in Theorem 2, the detailed .' is incomplete; it should refer to the appendix or be finished.
  2. [Definition 1 and Algorithm 1] In Definition 1, the notation H_k in the second branch should make the summation range explicit; the current 'Xk_{s=1}' is ambiguous and is repeated in Algorithm 1.
  3. [Table 1] The reported 'Improve' values for AttMix on Hetrec2011-ML include one negative value (recall@5, -6.25%) but the text says ALKDRec 'significantly outperforms ... in most cases'; this is acceptable, but the negative case should be acknowledged in the discussion.
  4. [Experiments and Analysis] The efficiency claim ('around 44 minutes and 8.6 USD' vs '1782 minutes and 347.0 USD') is useful but the exact inference setup, including the number of LLM calls for summarization versus ranking, should be reported in a table for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the max-min selection theorem is a conditional mathematical result, and the reported SBR improvements are measured on held-out test sets; the grid-searched 1:5:4 count ratio is a robustness caveat, not a circular derivation.

full rationale

The paper's central derivation is a max-min game: for fixed counts (kef, ksi, kin) and fixed gain values, Theorem 1 proves a lower bound for the expected gain under p*, Theorem 2 proves an upper bound for a constructed worst-case combination, and Theorem 3 combines these to identify the equilibrium policy p*. The proofs do not assume the conclusion; they algebraically evaluate z(p*, c) and z(p, c-hat). The active learning policy p* is therefore a genuine consequence of the stated optimization problem, not a restatement of the input. The empirical claim that ALKDRec outperforms KD baselines is grounded in held-out test metrics (recall@K, ndcg@K) computed over five runs with t-tests, so it is not forced by construction. The grid-searched 1:5:4 ratio for effective/similar/incorrect instances is an implementation assumption: the theoretical guarantee in Theorem 3 applies only when the assumed counts match the true counts. If the true counts differ on a new dataset, the lower bound may not hold in deployment. This is a real threat to external validity and a limitation of the paper, but it is not circularity: the theorem remains a valid conditional statement, and the final test-set comparisons are measured rather than derived from the assumed ratio. Citations to prior work, including Baykal et al. (2023), are used as a starting point for the max-min formulation and are not load-bearing self-citations; the authors' own previous papers are cited for background and do not supply the uniqueness or correctness of the result. No equation in the paper reduces to its own inputs by definition, and no fitted parameter is renamed as a prediction. Therefore the paper receives a circularity score of 0.

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

Most of the selection policy's behavior is governed by gain values and type counts that are either hand-picked or grid-searched rather than measured; the theorem's guarantee is conditional on those quantities being correct.

free parameters (5)
  • mu = 10
    Exponent in g^ef = 1/[rank(d_fs)]^mu mapping difficulty to effective gain; set empirically in Implementation Details.
  • kef:ksi:kin ratio = 1:5:4
    Grid-searched counts of effective/similar/incorrect instances required by the max-min formulation; set in Hyperparameter Study and Implementation Details.
  • g_si and g_in = g^ef / 2
    Gain values for similar and incorrect instances chosen by hand in Implementation Details, directly shaping the selection policy.
  • tau = 500 (750 for FPMC on Amazon-Games)
    Number of LLM-queried sessions, chosen empirically; see Figure 4 and Implementation Details.
  • alpha_v = 3/2/1 for ranking positions [1,5]/(5,15]/(15,25]
    Pair-wise loss weights in Equation (1), set by hand for distillation weighting.
assumptions (4)
  • domain assumption Known total counts of effective, similar, and incorrect instances before LLM inference.
    Theorem 3's solution assumes kef, ksi, kin are known; in practice they are tuned as hyperparameters, as noted in the Active Learning section.
  • domain assumption Difficulty is measured by negative consistency between session and item embeddings.
    The d_fs formula encodes the claim that low consistency means the conventional recommender cannot capture the knowledge, which is a modeling choice.
  • ad hoc to paper Gains for similar and incorrect predictions are half the effective gain.
    This choice has no external justification and directly determines p*, so it is load-bearing for the selection policy.
  • standard math The max-min theorem from Baykal et al. (2023) extends to this ranking setting with three indicator types.
    The paper adapts Baykal et al.'s robust active distillation; the proof sketch contains typos and undefined terms, so the bound should be checked before relying on it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Active Large Language Model-based Knowledge Distillation for Session-based Recommendation." pith.science (2026). https://pith.science/paper/FOJ2YY2S

@misc{pith2026250215685,
  author       = {Pith},
  title        = {Pith review of: Active Large Language Model-based Knowledge Distillation for Session-based Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FOJ2YY2S}},
  note         = {Machine review of arXiv:2502.15685}
}
read the original abstract

Large language models (LLMs) provide a promising way for accurate session-based recommendation (SBR), but they demand substantial computational time and memory. Knowledge distillation (KD)-based methods can alleviate these issues by transferring the knowledge to a small student, which trains a student based on the predictions of a cumbersome teacher. However, these methods encounter difficulties for \textit{LLM-based KD in SBR}. 1) It is expensive to make LLMs predict for all instances in KD. 2) LLMs may make ineffective predictions for some instances in KD, e.g., incorrect predictions for hard instances or similar predictions as existing recommenders for easy instances. In this paper, we propose an active LLM-based KD method in SBR, contributing to sustainable AI. To efficiently distill knowledge from LLMs with limited cost, we propose to extract a small proportion of instances predicted by LLMs. Meanwhile, for a more effective distillation, we propose an active learning strategy to extract instances that are as effective as possible for KD from a theoretical view. Specifically, we first formulate gains based on potential effects (e.g., effective, similar, and incorrect predictions by LLMs) and difficulties (e.g., easy or hard to fit) of instances for KD. Then, we propose to maximize the minimal gains of distillation to find the optimal selection policy for active learning, which can largely avoid extracting ineffective instances in KD. Experiments on real-world datasets show that our method significantly outperforms state-of-the-art methods for SBR.

Figures

Figures reproduced from arXiv: 2502.15685 by the authors.

Figure 1
Figure 1. The architecture of the ALKDRec method. Distill knowledge from the LLM teacher To distill knowledge from the LLM teacher into the student, we fine-tune the student recommender Srec(·) that is well￾trained based on all instances S, thus making the student recommender learn to imitate the prediction of the LLM teacher, i.e., rankllm s , with pair-wise loss, maxX s∈S X v∈rankllms αv · log σ(Srec(s, v) − Srec(s, v′ )), … view at source ↗
Figure 2
Figure 2. Performance (ndcg@10) with different (a) latent [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Performance (ndcg@10) of ALKDRec across var [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Performance (ndcg@5/10) of ALKDRec vary with [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. SmartGR: Hierarchy and Beam-Aware Knowledge Distillation for Generative Recommendation

    cs.IR 2026-08 conditional novelty 6.0 of 10

    SmartGR distills a large generative recommender into a smaller one with hierarchy-aware SID and beam-aware ranking losses, improving metrics by 8.6% on average while keeping the smaller model's speed.

Reference graph

Works this paper leans on

46 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [1]

    Baykal, C.; Trinh, K.; Iliopoulos, F.; Menghani, G.; and Vee, E. 2023. Robust Active Distillation. In The Eleventh International Conference on Learning Representations (ICLR)

  2. [2]

    Chen, G.; Chen, J.; Feng, F.; Zhou, S.; and He, X. 2023. Unbiased Knowledge Distillation for Recommendation. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining (WSDM), 976--984

  3. [3]

    Chen, X.; Zhang, Y.; Xu, H.; Qin, Z.; and Zha, H. 2018. Adversarial Distillation for Efficient Recommendation with External Knowledge. ACM Transactions on Information Systems (TOIS), 1--28

  4. [4]

    Cui, Y.; Liu, F.; Wang, P.; Wang, B.; Tang, H.; Wan, Y.; Wang, J.; and Chen, J. 2024. Distillation Matters: Empowering Sequential Recommenders to Match the Performance of Large Language Model. arXiv preprint arXiv:2405.00338

  5. [5]

    Dai, S.; Shao, N.; Zhao, H.; Yu, W.; Si, Z.; Xu, C.; Sun, Z.; Zhang, X.; and Xu, J. 2023. Uncovering ChatGPT’s Capabilities in Recommender Systems. In Proceedings of the 17th ACM Conference on Recommender Systems (RecSys), 1126--1132

  6. [6]

    Deng, Y.; Chen, Y.; Dong, X.; Pan, L.; Li, H.; Cheng, L.; and Mo, L. 2023. BKD: A Bridge-based Knowledge Distillation Method for Click-Through Rate Prediction. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 1859--1863

  7. [7]

    Du, Y.; Luo, D.; Yan, R.; Wang, X.; Liu, H.; Zhu, H.; Song, Y.; and Zhang, J. 2024. Enhancing Job Recommendation Through LLM-based Generative Adversarial Networks. In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 8363--8371

  8. [8]

    Gao, Y.; Xiong, Y.; Gao, X.; Jia, K.; Pan, J.; Bi, Y.; Dai, Y.; Sun, J.; and Wang, H. 2023. Retrieval-augmented Generation for Large Language Models: A Survey. arXiv preprint arXiv:2312.10997

Show all 46 references
  1. [9]

    Geng, S.; Liu, S.; Fu, Z.; Ge, Y.; and Zhang, Y. 2022. Recommendation as Language Processing (RLP): A Unified Pretrain, Personalized Prompt & Predict Paradigm (p5). In Proceedings of the 16th ACM Conference on Recommender Systems (RecSys), 299--315

  2. [10]

    Hou, Y.; Zhang, J.; Lin, Z.; Lu, H.; Xie, R.; McAuley, J.; and Zhao, W. X. 2024. Large Language Models are Zero-shot Rankers for Recommender Systems. In European Conference on Information Retrieval (ECIR), 364--381

  3. [11]

    Huang, F.; Wang, Z.; Huang, X.; Qian, Y.; Li, Z.; and Chen, H. 2023. Aligning Distillation for Cold-start Item Recommendation. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 1147--1157

  4. [12]

    Kang, S. 2024. Knowledge Distillation Approaches for Accurate and Efficient Recommender System. arXiv preprint arXiv:2407.13952

  5. [13]

    Kang, S.; Hwang, J.; Kweon, W.; and Yu, H. 2020. DE-RRD: A Knowledge Distillation Framework for Recommender System. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management (CIKM), 605--614

  6. [14]

    Kang, S.; Hwang, J.; Kweon, W.; and Yu, H. 2021. Topology Distillation for Recommender System. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining (KDD), 829--839

  7. [15]

    Kang, S.; Kweon, W.; Lee, D.; Lian, J.; Xie, X.; and Yu, H. 2023 a . Distillation from Heterogeneous Models for Top-K Recommendation. In Proceedings of the ACM Web Conference 2023, 801--811

  8. [16]

    Kang, W.-C.; Ni, J.; Mehta, N.; Sathiamoorthy, M.; Hong, L.; Chi, E.; and Cheng, D. Z. 2023 b . Do LLMs Understand User Preferences? Evaluating LLMs On User Rating Prediction. arXiv preprint arXiv:2305.06474

  9. [17]

    Kweon, W.; Kang, S.; and Yu, H. 2021. Bidirectional Distillation for Top-K Recommender System. In Proceedings of the Web Conference 2021, 3861--3871

  10. [18]

    Lee, J.-w.; Choi, M.; Lee, J.; and Shim, H. 2019. Collaborative Distillation for Top-N Recommendation. In 2019 IEEE International Conference on Data Mining (ICDM), 369--378

  11. [19]

    Lee, Y.; and Kim, K.-E. 2021. Dual Correction Strategy for Ranking Distillation in Top-N Recommender System. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management (CIKM), 3186--3190

  12. [20]

    Liu, C.; Xie, R.; Liu, X.; Wang, P.; Zheng, R.; Zhang, L.; Li, J.; Xia, F.; and Lin, L. 2023 a . Future Augmentation with Self-distillation in Recommendation. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (ECML/PKDD), 602--618. Springer

  13. [21]

    Liu, F.; Chen, H.; Cheng, Z.; Nie, L.; and Kankanhalli, M. 2023 b . Semantic-Guided Feature Distillation for Multimodal Recommendation. In Proceedings of the 31st ACM International Conference on Multimedia (MM), 6567--6575

  14. [22]

    Liu, Q.; Chen, N.; Sakai, T.; and Wu, X.-M. 2024 a . Once: Boosting Content-based Recommendation with Both Open-and Closed-source Large Language Models. In Proceedings of the 17th ACM International Conference on Web Search and Data Mining (WSDM), 452--461

  15. [23]

    Liu, Q.; Wu, X.; Zhao, X.; Zhu, Y.; Zhang, Z.; Tian, F.; and Zheng, Y. 2024 b . Large Language Model Distilling Medication Recommendation Model. arXiv preprint arXiv:2402.02803

  16. [24]

    Liu, Q.; Zeng, Y.; Mokhosi, R.; and Zhang, H. 2018. STAMP: Short-Term Attention/Memory Priority Model for Session-based Recommendation. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining (KDD), 1831--1839

  17. [25]

    Luo, S.; Yao, Y.; He, B.; Huang, Y.; Zhou, A.; Zhang, X.; Xiao, Y.; Zhan, M.; and Song, L. 2024. Integrating large language models into recommendation via mutual augmentation and adaptive aggregation. arXiv preprint arXiv:2401.13870

  18. [26]

    Pan, Y.; He, F.; and Yu, H. 2019. A novel Enhanced Collaborative Autoencoder with Knowledge Distillation for Top-N Recommender Systems. Neurocomputing, 332: 137--148

  19. [27]

    Ren, X.; Wei, W.; Xia, L.; Su, L.; Cheng, S.; Wang, J.; Yin, D.; and Huang, C. 2024. RLMRec: Representation Learning with Large Language Models for Recommendation. In Proceedings of the ACM on Web Conference 2024, 3464--3475

  20. [28]

    Rendle, S.; Freudenthaler, C.; and Schmidt-Thieme, L. 2010. Factorizing Personalized Markov Chains for Next-Basket Recommendation. In Proceedings of the 19th international conference on World wide web (WWW), 811--820

  21. [29]

    Shi, T.; Zhang, Y.; Xu, Z.; Chen, C.; Feng, F.; He, X.; and Tian, Q. 2023. Preliminary Study on Incremental Learning for Large Language Model-based Recommender Systems. arXiv preprint arXiv:2312.15599

  22. [30]

    X.; Lin, L.; and Wen, J.-R

    Sun, W.; Xie, R.; Zhang, J.; Zhao, W. X.; Lin, L.; and Wen, J.-R. 2024 a . Distillation is All You Need for Practically Using Different Pre-trained Recommendation Models. arXiv preprint arXiv:2401.00797

  23. [31]

    Sun, Z.; Fang, H.; Yang, J.; Qu, X.; Liu, H.; Yu, D.; Ong, Y.-S.; and Zhang, J. 2022. Daisyrec 2.0: Benchmarking Recommendation for Rigorous Evaluation. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 8206--8226

  24. [32]

    Sun, Z.; Liu, H.; Qu, X.; Feng, K.; Wang, Y.; and Ong, Y. S. 2024 b . Large Language Models for Intent-Driven Session Recommendations. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 324--334

  25. [33]

    Tang, J.; and Wang, K. 2018. Ranking Distillation: Ranking Distillation: Learning Compact Ranking Models With High Performance for Recommender System. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining (KDD), 2289--2298

  26. [34]

    Wang, S.; Sui, Y.; Wu, J.; Zheng, Z.; and Xiong, H. 2024 a . Dynamic Sparse Learning: A Novel Paradigm for Efficient Recommendation. In Proceedings of the 17th ACM International Conference on Web Search and Data Mining (WSDM), 740--749

  27. [35]

    Wang, S.; Zhang, K.; Wu, L.; Ma, H.; Hong, R.; and Wang, M. 2021. Privileged Graph Distillation for Cold Start Recommendation. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 1187--1196

  28. [36]

    Wang, X.; Zhang, R.; Sun, Y.; and Qi, J. 2018. KDGAN: Knowledge Distillation with Generative Adversarial Networks. Advances in neural information processing systems (NeurIPS)

  29. [37]

    Wang, Y.; Tian, C.; Hu, B.; Yu, Y.; Liu, Z.; Zhang, Z.; Zhou, J.; Pang, L.; and Wang, X. 2024 b . Can Small Language Models be Good Reasoners for Sequential Recommendation? In Proceedings of the ACM on Web Conference 2024, 3876--3887

  30. [38]

    Wei, W.; Ren, X.; Tang, J.; Wang, Q.; Su, L.; Cheng, S.; Wang, J.; Yin, D.; and Huang, C. 2024 a . LLMRec: Large Language Models with Graph Augmentation for Recommendation. In Proceedings of the 17th ACM International Conference on Web Search and Data Mining (WSDM), 806--815

  31. [39]

    Wei, W.; Tang, J.; Xia, L.; Jiang, Y.; and Huang, C. 2024 b . Multi-Modal Knowledge Distillation for Recommendation with Prompt-Tuning. In The Web Conference, 3217–3228

  32. [40]

    Wu, L.; Zheng, Z.; Qiu, Z.; Wang, H.; Gu, H.; Shen, T.; Qin, C.; Zhu, C.; Zhu, H.; Liu, Q.; et al. 2023. A Survey on Large Language Models for Recommendation. arXiv preprint arXiv:2305.19860

  33. [41]

    Xia, X.; Yin, H.; Yu, J.; Wang, Q.; Xu, G.; and Nguyen, Q. V. H. 2022. On-Device Next-Item Recommendation with Self-Supervised Knowledge Distillation. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 546--555

  34. [42]

    Xu, C.; Li, Q.; Ge, J.; Gao, J.; Yang, X.; Pei, C.; Sun, F.; Wu, J.; Sun, H.; and Ou, W. 2020. Privileged Features Distillation at Taobao Recommendations. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD), 2590--2598

  35. [43]

    B.; Zhang, Y.; Xie, X.; Wang, H.; and Kim, S

    Zhang, P.; Guo, J.; Li, C.; Xie, Y.; Kim, J. B.; Zhang, Y.; Xie, X.; Wang, H.; and Kim, S. 2023. Efficiently Leveraging Multi-level User Intent for Session-based Recommendation via Atten-Mixer Network. In Proceedings of the sixteenth ACM international conference on web search ...

  36. [44]

    Zhang, Y.; Xu, X.; Zhou, H.; and Zhang, Y. 2020. Distilling Structured Knowledge into Embeddings for Explainable and Accurate Recommendation. In Proceedings of the 13th international conference on web search and data mining (WSDM), 735--743

  37. [45]

    X.; Zhou, K.; Li, J.; Tang, T.; Wang, X.; Hou, Y.; Min, Y.; Zhang, B.; Zhang, J.; Dong, Z.; et al

    Zhao, W. X.; Zhou, K.; Li, J.; Tang, T.; Wang, X.; Hou, Y.; Min, Y.; Zhang, B.; Zhang, J.; Dong, Z.; et al. 2023. A Survey of Large Language Models. arXiv preprint arXiv:2303.18223

  38. [46]

    Zheng, Z.; Qiu, Z.; Hu, X.; Wu, L.; Zhu, H.; and Xiong, H. 2023. Generative Job Recommendations with Large Language Model. arXiv preprint arXiv:2307.02157

Pith tools

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