Pith. sign in

REVIEW 4 major objections 5 minor 38 references

Bridging the Gap: Self-Optimized Fine-Tuning for LLM-based Recommender Systems

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A 'Guidance+Tuning' curriculum that has an LLM recommender learn from its own self-distilled, easy-to-learn outputs before facing real user data improves recommendation accuracy by 37.59% on average.

desk verdict Plausible and clearly-described curriculum recipe for LLM recommenders, but the headline 37.59% gain lacks statistical support and the paper needs variance reporting and a sharper baseline comparison before the claim is credible. read the letter →

arxiv 2505.20771 v1 pith:IEGS3BPM submitted 2025-05-27 cs.IR cs.AI

Heng Tang , Feng Liu , Xinbo Chen , Jiawei Chen , Bohao Wang , Changwang Zhang , Jun Wang , Yuegang Sun
show 2 more authors
Bingde Hu Can Wang
This is my paper
classification cs.IRcs.AI
keywords LLM-basedrecommendersystemsself-distillationcurriculumlearningsupervisedfine-tuningsequentialrecommendationself-adaptiveschedulerknowledgegapaccuracy
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

Large language models know a great deal about language, but their knowledge space is misaligned with the structure of recommendation data, and the paper argues this gap is why plain prompting and plain supervised fine-tuning both fall short. To bridge the gap, the paper proposes Self-Optimized Fine-Tuning (SOFT), a 'Guidance+Tuning' strategy that trains an LLM in two stages: first on an auxiliary dataset built from the model's own predictions after fine-tuning (easy, self-distilled data), then progressively on the real recommendation data (hard data) under an adaptive curriculum schedule. Across three Amazon datasets and two LLM backbones, SOFT reports an average accuracy gain of 37.59% over the corresponding supervised fine-tuning baselines. The paper confines this claim to sequential recommendation over named items with LoRA-based fine-tuning, as stated in its Limitations section.

What carries the argument

The load-bearing objects are the self-distilled auxiliary dataset $\hat{\mathcal{D}}$ and the self-adaptive curriculum scheduler $\tau$. Self-distillation constructs $\hat{\mathcal{D}}$ by pairing each prompt with the output of the already fine-tuned LLM, producing data that lies within the model's own distribution and carries the coarse category-level recommendation knowledge the model has managed to learn. The scheduler $\tau=e^{\alpha(d_t/d_0-1)}$ balances the distilled-data loss against the real-data loss, with $d_t$ the average L2 distance in token-embedding space between generated outputs and true target items; as training proceeds and $d_t$ falls, weight shifts from easy self-generated data to hard real data. Together these two components turn supervised fine-tuning into a two-stage curriculum: first exploit the model's own easy knowledge space, then tune it into the external hard knowledge space.

What would settle it

Run SOFT against a control in which each self-distilled item name is replaced by a random item from the same category; if the control reproduces the gains, the easy distribution, not the semantic content, is what transfers. Separately, give standard SFT on real data enough additional epochs to match SOFT's total gradient updates and check whether the reported 37.59% gap closes to zero.

Watch

Extended reading notes

Core claim

The paper's central claim is that the knowledge gap between an LLM and a recommendation task is bridged by a curriculum built from the model's own outputs. After supervised fine-tuning, the LLM generates a predicted next item for each training prompt; these predictions are seldom the exact target item (Hit Ratio@1 below 1%), but they land in the correct item category about 44.6% of the time on average, and the first-epoch loss on the self-distilled data is only 47% of the loss on the real data. SOFT therefore treats the self-distilled set as the easy stage and mixes it with the real dataset through $\mathcal{L}_{SOFT}=(1-\tau)\mathcal{L}_{SFT}+\tau\mathcal{L}_{SDFT}$, where the scheduler $\tau=e^{\alpha(d_t/d_0-1)}$ drops automatically as the embedding-space distance between the model's generated outputs and the target items shrinks. The reported result is that this easy-to-hard progression improves both BIGRec and LLaRA on Video Games, Movies and TV, and Kindle Store, with a 37.59% average improvement over SFT, and that the full method beats both training on the self-distilled data alone and training on the real data alone.

Load-bearing premise

The load-bearing premise is that the self-distilled data, which consists of the fine-tuned model's own often-wrong predicted item names, is easy to learn and carries the category-level recommendation knowledge that transfers to later learning on real data, rather than merely reinforcing hallucinations or adding extra training iterations.

Editorial extensions

If this is right

  • Because SOFT acts only at the loss level, any SFT-trained LLM recommender can adopt it without changing prompts, item vocabularies, or the grounding step.
  • Training on the self-distilled dataset alone already beats SFT on most settings, indicating that the model's own outputs filter out the hardest real-data knowledge and provide a useful warm start.
  • The full SOFT combination outperforms both guidance-only and tuning-only ablations, locating the benefit in the curriculum's mixing rather than in either loss term by itself.
  • Since LLM training converges within a few epochs, the exact functional form of the scheduler decay matters little; the advantage comes from starting the descent in a better region of parameter space, not from the shape of the $\tau$ curve.
  • The gains replicate across two backbones and three datasets, suggesting the loss-level curriculum transfers better than adding ID-based item embeddings to the input.

Reading between the lines

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

  • Beyond the paper: if the 44.6% category-level accuracy is the real carrier of the curriculum, SOFT should improve category-level hit rates earlier and more strongly than item-level hit rates; the reported metrics do not separate these, so this is a testable prediction rather than a paper claim.
  • Beyond the paper: the same loss-mixing recipe could transfer to other LLM fine-tuning tasks with a large distribution gap, such as structured or domain-specific generation, where a model's own coarse outputs could serve as an easy scaffold; the paper does not claim this.
  • Beyond the paper: running SOFT with a control self-distilled dataset whose predicted item names are replaced by random same-category names would isolate whether the transfer comes from the easy-to-learn distribution or from the semantic category content, since the paper's evidence correlates but does not separate these two factors.
  • Beyond the paper: the paper's observation that SDPO underperforms SFT in this all-ranking setting suggests the curriculum may also be regularizing token distributions; combining SOFT with preference optimization is an unstated extension.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Self-Optimized Fine-Tuning (SOFT), a curriculum-learning strategy for LLM-based recommender systems. SOFT first constructs an auxiliary dataset by using an SFT-finetuned LLM to generate its own output sequences (self-distillation), then fine-tunes the LLM on a weighted mixture of this self-distilled data and the real recommendation data, with a self-adaptive scheduler that adjusts the mixing weight τ based on the embedding distance between the model's current outputs and the target items. Experiments on three Amazon datasets (Game, Movie, Kindle) with two LLM backbones (BIGRec, LLaRA) report that SOFT improves over SFT and SDPO baselines by an average of 37.59% across metrics, and the paper includes an ablation of the scheduler and a hyperparameter sensitivity study.

Significance. If the empirical claim is validated, the idea of using a model's own outputs as an easy-to-learn curriculum before exposing it to real recommendation data is a simple and potentially useful contribution to LLM-based recommendation, and the paper provides a clear loss formulation and an open-source implementation. The paper also offers a plausible mechanistic story (lower first-epoch loss on self-distilled data, category-level knowledge) and an ablation showing that the scheduler contributes to the gains. However, the central claim of "significant" improvement currently rests on averages over five runs with no variance reporting, no significance tests, and only 1,000 test sequences, so the magnitude of the reported gains is not yet established. The mechanism is also not isolated from confounds such as the total amount of training data or the effect of training on the model's own outputs as a regularizer.

major comments (4)
  1. [§4.1, Table 2] The claim that SOFT "significantly enhances the recommendation accuracy (37.59% on average)" is not statistically supported. Table 2 reports only averages over five runs without standard deviations, confidence intervals, or significance tests. Several individual improvements are very small in absolute terms: for example, LLaRA on Movie H@20 is 0.0200 (SFT) versus 0.0210 (SOFT), a difference of 0.001, which is one hit in the 1,000-sequence test set, and LLaRA on Movie H@5 is 0.0170 versus 0.0200, a difference of three hits. Given the stochasticity of LoRA fine-tuning and the modest training set size (4,096 sequences), these differences may be within run-to-run noise. Please report per-seed results or standard deviations, and provide paired significance tests (e.g., bootstrap or Wilcoxon signed-rank) over the five runs for each metric and dataset.
  2. [§4.1, §4.4, Figure 5] The hyperparameter α is tuned over {0.1, 1, 10, 100} for each dataset and backbone, but the paper does not state whether this selection is performed on the validation set or on the test set. Figure 5 shows that performance varies substantially with α, so if test performance was used to select α, the reported gains are inflated by selection bias. Please clarify the model selection protocol: state which split was used to choose α, report the chosen α values for each setting, and, if validation-based selection was used, give the corresponding validation metric.
  3. [§3.1, §4.3, Figure 2] The mechanism claim—that self-distilled data provides an easy-to-learn but meaningful curriculum that improves later learning on real data—is not isolated by the experiments. The lower first-epoch loss on the SD dataset (Figure 2a) is expected when a model is trained to imitate its own outputs, and the 44.6% category-level accuracy (Figure 2b) does not by itself show that this knowledge transfers to better real-data performance. The ablation "w/o SA" trains only on SD data, which does not control for the total number of gradient steps, the amount of data seen per step, or the effect of training on the model's own outputs as a regularizer. Please add control conditions that separate these factors, such as (i) a version of SOFT that mixes real data with a matched amount of randomly selected or synthetic auxiliary data, (ii) a curriculum using SD data generated by a different model or from an earlier checkpoint, and (iii) a fixed-τ (non-adaptive) curriculum with the same data mixture, to attribute the gains to the proposed SD curriculum rather than to the scheduler alone or to the additional training signal.
  4. [§4.1] The evaluation uses only 1,000 test sequences and 4,096 training sequences per dataset. With the all-ranking protocol, the variance of H@5 and H@20 on 1,000 sequences is substantial; for example, a difference of 0.001 corresponds to a single hit. The paper should report confidence intervals (e.g., bootstrap over test sequences) and should either justify the sample sizes or increase them. This is particularly important given that the reported absolute gains are often only a few hits, which may not be practically meaningful even if statistically significant.
minor comments (5)
  1. [Table 2 caption] The caption states that "Gain" is the improvement over "other fine-tuning strategies," but it does not specify that the baseline is the better of SFT and SDPO, not SFT alone. Please make this explicit, as the percentage gains are relative to the best fine-tuning baseline.
  2. [§3.1] The paper says the self-distilled dataset is generated by "LLMs after SFT" but does not explicitly state that the SD dataset is fixed before the curriculum training begins and is not regenerated during SOFT training. Please state this clearly, as it affects the interpretation of the self-adaptive scheduler.
  3. [§2.2 and §3.2] The definition of the distance d_t uses token embeddings z_y, but the paper does not explain how the generated output y_t is grounded to an item name or description. Please specify the grounding procedure, since the distance calculation and the SD data construction both depend on it.
  4. [§4.1] There are several typos: "validing" should be "validation" in the dataset description, "searving" should be "serving" in Section 4.5, and "incorperates" should be "incorporates" in the baseline description.
  5. [§3.3] In the sentence "This process has been demonstrated to be equivalent to unsupervised pre-training, which allows the training to to start in better basins of attraction," there is a duplicated "to". Please fix this typo.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: self-distillation is self-referential by design, but the headline result is measured on held-out real items.

full rationale

The paper's derivation chain is not circular. SOFT constructs the auxiliary SD dataset by sampling outputs of the SFT model (Section 3.1: 'ˆyi = LLMSF T(xi)') and then optimizes L_SOFT = (1 − τ)L_SFT + τ L_SDFT, with τ driven by the model's own output-to-item distance (Section 3.2). This is self-referential in mechanism, but the claimed result—higher H@K and NG@K on held-out real items (Table 2)—is evaluated against real next-item labels, not against the SD labels or the scheduler's own distance. The SD data and scheduler are training inputs, not the evaluation target; the improvement over SFT is an empirical outcome that could in principle fail. The paper's evidence for the 'easy and meaningful' property of the SD dataset (Figure 2: first-epoch loss ratio and 44.6% category-level accuracy) is observational motivation, not a definitional equivalence that forces the reported gains. The self-citations to DROS and DLLM2Rec are baseline implementations, not load-bearing premises for the central claim. The appended Limitations section restricts the scope to sequential recommendation and LoRA fine-tuning but does not concede any circular step. The absence of error bars or significance tests is a statistical-reporting weakness and a correctness risk, not a circularity, because it does not make any prediction equivalent to its inputs by construction. No step reduces to its own inputs, so the appropriate score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the self-distilled dataset being genuinely easier and yet meaningful, and on the scheduler's distance measure being a reliable gauge of learning progress. These are domain assumptions, not established results. The free parameters α and M are hand-chosen or tuned, while the loss functions themselves are standard SFT objectives.

free parameters (2)
  • alpha (curriculum scheduler exponent) = tuned per dataset/backbone from {0.1, 1, 10, 100}
    Controls how quickly the training shifts from self-distilled data to real data. The paper reports that the optimal value differs across datasets and backbones, so it is selected on validation performance, which is a free parameter fitted to the data.
  • M (number of samples for distance estimation) = 256
    The number of training samples used to estimate the average embedding distance d_t. The paper sets this to 256 without a sensitivity study, making it a hand-chosen hyperparameter that affects the scheduler's behavior.
assumptions (4)
  • domain assumption Curriculum learning improves optimization by starting from easier data and shifting to harder data (Bengio et al., 2009).
    The method's rationale rests on the claim that training on easy data first leads to better basins of attraction. This is a machine-learning prior, not a formal theorem, and is invoked in Section 3.3.
  • domain assumption The self-distilled dataset generated by the fine-tuned model contains meaningful, easy-to-learn recommendation knowledge (not just noise).
    Section 3.1 claims the SD data is easier (lower first-epoch loss) and category-level accurate (44.6% category accuracy), but this is an empirical assertion, not a proven property. The method's success depends on this.
  • domain assumption The L2 distance in token embedding space between generated and target item descriptions is a valid measure of how far the model is from the correct recommendation.
    Used in Section 3.2 for the scheduler and in the grounding step from Bao et al. (2023). The paper assumes embedding distance reflects recommendation competence.
  • standard math Cross-entropy supervised fine-tuning is the appropriate training objective for LLM-based recommenders.
    The SFT loss in Section 2.2 is the objective the paper builds on; this is a standard but domain-specific choice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging the Gap: Self-Optimized Fine-Tuning for LLM-based Recommender Systems." pith.science (2026). https://pith.science/paper/IEGS3BPM

@misc{pith2026250520771,
  author       = {Pith},
  title        = {Pith review of: Bridging the Gap: Self-Optimized Fine-Tuning for LLM-based Recommender Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IEGS3BPM}},
  note         = {Machine review of arXiv:2505.20771}
}
read the original abstract

Recent years have witnessed extensive exploration of Large Language Models (LLMs) on the field of Recommender Systems (RS). There are currently two commonly used strategies to enable LLMs to have recommendation capabilities: 1) The "Guidance-Only" strategy uses in-context learning to exploit and amplify the inherent semantic understanding and item recommendation capabilities of LLMs; 2) The "Tuning-Only" strategy uses supervised fine-tuning (SFT) to fine-tune LLMs with the aim of fitting them to real recommendation data. However, neither of these strategies can effectively bridge the gap between the knowledge space of LLMs and recommendation, and their performance do not meet our expectations. To better enable LLMs to learn recommendation knowledge, we combine the advantages of the above two strategies and proposed a novel "Guidance+Tuning" method called Self-Optimized Fine-Tuning (SOFT), which adopts the idea of curriculum learning. It first employs self-distillation to construct an auxiliary easy-to-learn but meaningful dataset from a fine-tuned LLM. Then it further utilizes a self-adaptive curriculum scheduler to enable LLMs to gradually learn from simpler data (self-distilled data) to more challenging data (real RS data). Extensive experiments demonstrate that SOFT significantly enhances the recommendation accuracy (37.59\% on average) of LLM-based methods. The code is available via https://anonymous.4open.science/r/Self-Optimized-Fine-Tuning-264E

Figures

Figures reproduced from arXiv: 2505.20771 by the authors.

Figure 1
Figure 1. Three different training strategies used in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (left) Comparison of the training loss on the [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. The comparison of different training proce [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (1 more)
Figure 6
Figure 6. Figure 6: The training time (minutes) and performance [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 25 canonical work pages

  1. [1]

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

    arXiv 2023

  2. [2]

    Yoshua Bengio, J \'e r \^o me Louradour, Ronan Collobert, and Jason Weston. 2009. Curriculum learning. In ICML, pages 41--48

    work page 2009

  3. [3]

    Jianxin Chang, Chen Gao, Yu Zheng, Yiqun Hui, Yanan Niu, Yang Song, Depeng Jin, and Yong Li. 2021. Sequential recommendation with graph neural networks. In Proceedings of the 44th international ACM SIGIR conference on research and development in information retrieval, pages 378--387

    work page 2021

  4. [4]

    Huayu Chen, Guande He, Lifan Yuan, Ganqu Cui, Hang Su, and Jun Zhu. 2024 a . Noise contrastive alignment of language models with explicit rewards. arXiv preprint arXiv:2402.05369

    arXiv 2024

  5. [5]

    Yuxin Chen, Junfei Tan, An Zhang, Zhengyi Yang, Leheng Sheng, Enzhi Zhang, Xiang Wang, and Tat-Seng Chua. 2024 b . On softmax direct preference optimization for recommendation. arXiv preprint arXiv:2406.09215

    arXiv 2024

  6. [6]

    Yu Cui, Feng Liu, Pengbo Wang, Bohao Wang, Heng Tang, Yi Wan, Jun Wang, and Jiawei Chen. 2024. Distillation matters: empowering sequential recommenders to match the performance of large language models. In RecSys, pages 507--517

    work page 2024

  7. [7]

    Tommaso Furlanello, Zachary Lipton, Michael Tschannen, Laurent Itti, and Anima Anandkumar. 2018. Born again neural networks. In ICML, pages 1607--1616

    work page 2018

  8. [8]

    Yunfan Gao, Tao Sheng, Youlin Xiang, Yun Xiong, Haofen Wang, and Jiawei Zhang. 2023. Chat-rec: Towards interactive and explainable llms-augmented recommender system. arXiv preprint arXiv:2303.14524

    arXiv 2023

Show all 38 references
  1. [9]

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783

    2024 arXiv

  2. [10]

    B Hidasi. 2015. Session-based recommendations with recurrent neural networks. arXiv preprint arXiv:1511.06939

    2015 arXiv

  3. [11]

    Yupeng Hou, Junjie Zhang, Zihan Lin, Hongyu Lu, Ruobing Xie, Julian McAuley, and Wayne Xin Zhao. 2024. Large language models are zero-shot rankers for recommender systems. In ECIR, page 364–381

    2024

  4. [12]

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685

    2021 arXiv

  5. [13]

    Jiaxin Huang, Shixiang Shane Gu, Le Hou, Yuexin Wu, Xuezhi Wang, Hongkun Yu, and Jiawei Han. 2022. Large language models can self-improve. arXiv preprint arXiv:2210.11610

    2022 arXiv

  6. [14]

    Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recommendation. In ICDM, pages 197--206. IEEE

    2018

  7. [15]

    Sein Kim, Hongseok Kang, Seungyoon Choi, Donghyun Kim, Minchul Yang, and Chanyoung Park. 2024. Large language models meet collaborative filtering: An efficient all-round llm-based recommender system. In SIGKDD, pages 1395--1406

    2024

  8. [16]

    Lei Li, Yongfeng Zhang, and Li Chen. 2023. Prompt distillation for efficient llm-based recommendation. In CIKM, page 1348–1357

    2023

  9. [17]

    Jiayi Liao, Sihang Li, Zhengyi Yang, Jiancan Wu, Yancheng Yuan, Xiang Wang, and Xiangnan He. 2024. Llara: Large language-recommendation assistant. In SIGIR, pages 1785--1795

    2024

  10. [18]

    Xinyu Lin, Wenjie Wang, Yongqi Li, Shuo Yang, Fuli Feng, Yinwei Wei, and Tat-Seng Chua. 2024. Data-efficient fine-tuning for llm-based recommendation. In SIGIR, pages 365--374

    2024

  11. [19]

    Qijiong Liu, Nuo Chen, Tetsuya Sakai, and Xiao-Ming Wu. 2024. Once: Boosting content-based recommendation with both open- and closed-source large language models. In WSDM, page 452–461

    2024

  12. [20]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2024. Direct preference optimization: Your language model is secretly a reward model. In NeurIPS, volume 36

    2024

  13. [21]

    Xubin Ren, Wei Wei, Lianghao Xia, Lixin Su, Suqi Cheng, Junfeng Wang, Dawei Yin, and Chao Huang. 2024. Representation learning with large language models for recommendation. In WWW, pages 3464--3475

    2024

  14. [22]

    Wentao Shi, Xiangnan He, Yang Zhang, Chongming Gao, Xinyue Li, Jizhi Zhang, Qifan Wang, and Fuli Feng. 2024. Large language models are learnable planners for long-term recommendation. In SIGIR, page 1893–1903

    2024

  15. [23]

    Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang. 2019. Bert4rec: Sequential recommendation with bidirectional encoder representations from transformer. In CIKM, pages 1441--1450

    2019

  16. [24]

    Zhiqing Sun, Yikang Shen, Qinhong Zhou, Hongxin Zhang, Zhenfang Chen, David Cox, Yiming Yang, and Chuang Gan. 2024 a . Principle-driven self-alignment of language models from scratch with minimal human supervision. In NeurIPS, volume 36

    2024

  17. [25]

    Zhongxiang Sun, Zihua Si, Xiaoxue Zang, Kai Zheng, Yang Song, Xiao Zhang, and Jun Xu. 2024 b . Large language models enhanced collaborative filtering. In CIKM, pages 2178--2188

    2024

  18. [26]

    Jiaxi Tang and Ke Wang. 2018. Personalized top-n sequential recommendation via convolutional sequence embedding. In WSDM, pages 565--573

    2018

  19. [27]

    Lei Wang and Ee-Peng Lim. 2023. Zero-shot next-item recommendation using large pretrained language models. arXiv preprint arXiv:2304.03153

    2023 arXiv

  20. [28]

    Xin Wang, Yudong Chen, and Wenwu Zhu. 2021. A survey on curriculum learning. In TPAMI, volume 44, pages 4555--4576. IEEE

    2021

  21. [29]

    Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2022. Self-instruct: Aligning language models with self-generated instructions. arXiv preprint arXiv:2212.10560

    2022 arXiv

  22. [30]

    Wei Wei, Xubin Ren, Jiabin Tang, Qinyong Wang, Lixin Su, Suqi Cheng, Junfeng Wang, Dawei Yin, and Chao Huang. 2024. Llmrec: Large language models with graph augmentation for recommendation. In WSDM, pages 806--815

    2024

  23. [31]

    Yunjia Xi, Weiwen Liu, Jianghao Lin, Xiaoling Cai, Hong Zhu, Jieming Zhu, Bo Chen, Ruiming Tang, Weinan Zhang, and Yong Yu. 2024. Towards open-world recommendation with knowledge augmentation from large language models. In RecSys, pages 12--22

    2024

  24. [32]

    Shenghao Yang, Weizhi Ma, Peijie Sun, Qingyao Ai, Yiqun Liu, Mingchen Cai, and Min Zhang. 2024 a . Sequential recommendation with latent relations based on large language model. In SIGIR, pages 335--344

    2024

  25. [33]

    Zhaorui Yang, Tianyu Pang, Haozhe Feng, Han Wang, Wei Chen, Minfeng Zhu, and Qian Liu. 2024 b . Self-distillation bridges distribution gap in language model fine-tuning. arXiv preprint arXiv:2402.13669

    2024 arXiv

  26. [34]

    Zhengyi Yang, Xiangnan He, Jizhi Zhang, Jiancan Wu, Xin Xin, Jiawei Chen, and Xiang Wang. 2023. A generic learning framework for sequential recommendation with distribution shifts. In SIGIR, pages 331--340

    2023

  27. [35]

    Eric Zelikman, YH Wu, Jesse Mu, and Noah D Goodman. 2024. Star: Self-taught reasoner bootstrapping reasoning with reasoning. In NeurIPS, volume 1126

    2024

  28. [36]

    Junjie Zhang, Ruobing Xie, Yupeng Hou, Xin Zhao, Leyu Lin, and Ji-Rong Wen. 2024. Recommendation as instruction following: A large language model empowered recommendation approach. In TOIS

    2024

  29. [37]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  30. [38]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.