Pith. sign in

REVIEW 5 major objections 6 minor 61 references

LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation

T0 review · 5 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read LaRec, a two-stage training framework, claims that LLM-based generative recommenders can reason inside a continuous latent space and thereby beat explicit chain-of-thought methods in accuracy while keeping inference latency nearly as low…

desk verdict A well-engineered latent-reasoning recipe for LLM recommenders, but the target-aware teacher may mean the 'reasoning' is answer distillation in disguise. read the letter →

arxiv 2607.24617 v1 pith:CX77ET4T submitted 2026-07-27 cs.IR

Yu Xia , Zihan Lin , Wei Yang , Rui Zhong , Cheng Chen , Huan Ren , Yao Hu This is my paper · ORCID
classification cs.IR
keywords latentreasoninggenerativerecommendationchain-of-thoughtdistillationreinforcementlearningGRPOGaussianmixtureexplorationsequentialLLMinferencelatency
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

The paper proposes LaRec, a training framework that makes LLM-based generative recommendation 'think' in a continuous latent space instead of emitting explicit chain-of-thought text. It claims that two added training stages—latent pre-training that distills fine-grained reasoning states from an explicit teacher, and personalized reinforcement tuning that samples user-specific starting points—let the model beat all tested baselines on four datasets while keeping inference latency nearly as low as a model that does no reasoning at all. If this holds, recommender systems could get the accuracy benefit of reasoning without the response-time cost that made explicit CoT difficult to deploy.

What carries the argument

The load-bearing mechanism is a pair of alignment losses inserted between the model's reasoning head and the text embedding space. Step-level alignment uses an InfoNCE contrastive loss that pulls each latent state $z_{i,k}$ toward the hidden state $h_{i,k}$ of the corresponding explicit CoT step within a mini-batch while pushing it away from other samples' states, giving every hidden step its own learning signal. Process direction alignment adds a hinge loss that forces the difference vector between consecutive latent states $\mathbf{v}_k = z_k - z_{k-1}$ to be cosine-similar to the scaled target item embedding $\frac{1}{K-1} e_{\mathrm{tgt}}$, thereby preventing 'semantic idling' in the latent space. A lightweight reasoning head maps the latent states into the text embedding space so that these alignment losses and the subsequent RL can operate on a shared representation.

What would settle it

Train a linear probe on the teacher's per-step hidden states (and on LaRec's latent states) to predict the ground-truth item from $z_k$ alone, conditioning out user-history information; if probe accuracy is substantially above chance and above the accuracy obtained when the same teacher is prompted without the target item, the 'no leakage' claim is falsified.

Watch

Extended reading notes

Core claim

LaRec's central claim is that latent reasoning can be made to work for sequential recommendation if the hidden reasoning states receive dense supervision and if the reasoning path is allowed to vary per user. The paper argues that earlier latent-reasoning attempts failed because they optimized the black-box states only through the final prediction loss and followed a single deterministic trajectory. LaRec instead (1) aligns each latent step $z_k$ with the hidden state of the corresponding explicit CoT step distilled from a large teacher LLM, (2) constrains each step's update vector $\mathbf{v}_k = z_k - z_{k-1}$ to point toward the target item's embedding, and (3) applies GRPO with rewards for exact hits and semantic similarity, sampling exploration start points from a per-user Gaussian mixture built from the user's historical item embeddings. Empirically the paper reports state-of-the-art H@5/H@10 and N@5/N@10 across the Toys, Instruments, Movie, and Industry datasets, with per-sample latency of 0.24–0.67 s versus 2.49–3.92 s for explicit CoT.

Load-bearing premise

The teacher's step-by-step reasoning is generated while the ground-truth next item is visible in the prompt, and the paper assumes the teacher's hidden states do not encode that item's identity even though the text is forbidden from naming it; if the hidden states leak the answer, the claimed reasoning capability is actually answer distillation.

Editorial extensions

If this is right

  • Latent reasoning can match or exceed explicit CoT's recommendation accuracy while keeping per-request latency close to models that do no reasoning at all.
  • Dense supervision of intermediate latent states, not just final-label feedback, is necessary to unlock latent reasoning; removing either alignment loss degrades results on all four datasets.
  • Anchored exploration via a user-specific Gaussian mixture outperforms both deterministic latent reasoning and random Gaussian-noise exploration in sparse recommendation spaces.
  • Latent reasoning exhibits LLM scaling laws: larger backbones improve H@10 and N@10 monotonically in the tested range from 0.5B to 7B parameters.
  • The framework transfers from public benchmarks to an industrial dataset and an online A/B test, where the paper reports a +2.93% conversion gain.

Reading between the lines

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

  • Because the process direction loss feeds the target item's embedding directly into the optimization, part of the reported gain may come from an answer-specific shortcut rather than from general reasoning; ablating this loss while substituting a target-free direction regularizer would isolate that contribution.
  • The teacher hidden states used for Step-level Alignment come from a model that saw the target item in the prompt; a truly blind history-only teacher would test whether the distilled states carry transferable preference logic or merely item identity.
  • The per-user Gaussian mixture sampler could serve as a controllable diversity knob at inference time: adjusting the sampled start point or the exploration radius $\sigma$ would let a deployed system trade accuracy for variety in recommendations.
  • The same two-stage recipe—distill explicit reasoning into latent states, then align with RL under a user-conditioned prior—could transfer to other generation tasks where verbose intermediate reasoning is expensive, such as conversational search or personalized summarization.
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

5 major / 6 minor

Summary. The paper proposes LaRec, a two-stage framework for latent reasoning in LLM-based generative recommendation. Stage one, Latent Pre-training (LPT), adds two auxiliary losses to the standard next-item loss: a step-level alignment loss (Eq. 4) that pulls latent reasoning states toward hidden states of an explicit chain-of-thought generated by Qwen-72B, and a process direction alignment loss (Eq. 6) that pushes each latent update toward the target item's mean-pooled embedding. Stage two, Personalized RL-tuning (PRL), injects a user-specific Gaussian mixture perturbation into the initial latent state and trains with GRPO under a composite hit-plus-semantic-similarity reward. Experiments are reported on Toys, Instruments, MovieLens-100K, and an industrial dataset, with claims of significant gains over all baselines, competitive latency, and a successful online A/B test.

Significance. If the mechanism claim held, the paper would make a useful contribution: it addresses two real limitations of latent reasoning for recommendation, namely sparse supervision of intermediate latents and deterministic reasoning paths, and it offers an efficiency argument relative to explicit CoT. The internal ablation results are broadly consistent with the design, the framework is clearly described, and the inclusion of an industrial deployment gives practical credibility. However, the central claim that LaRec 'unleashes latent reasoning' is currently not established because the teacher signals and the process-direction loss both contain the ground-truth item, so the reported gains are consistent with answer distillation. The paper also ships no code, no error bars for ablations, and no leakage probe, which weakens the empirical case for the specific reasoning mechanism.

major comments (5)
  1. [Section 3.2.1, Eq. (3), and Figure 3] The teacher hidden states h_k are computed from prompt_cot(S_u, i_tgt), where the ground-truth item title appears in the prompt context. Only the generated CoT text is prohibited from naming the target item; transformer hidden states attending to the prompt can encode i_tgt directly. The step-level alignment loss (Eq. 4) then trains z_k to be close to h_k, so the latent states may be optimized to contain answer information rather than to perform preference reasoning. Section 4.4.6 concludes that Target-Oriented CoT improves results 'without information leakage', but the supporting evidence is only the surface-text restriction. The authors should test whether h_k or z_k permit decoding of i_tgt, e.g., by training a linear probe on the teacher/latent states, and should compare against a teacher generated without i_tgt in the prompt or with the target removed from the conditioning context.
  2. [Section 3.2.2, Eq. (6)] The process direction alignment loss explicitly regresses each latent update v_k toward the target item embedding e_tgt, scaled by 1/(K-1). This is a second, direct path by which the ground-truth item enters the latent states. Even if the teacher CoT leakage were resolved, Eq. (6) alone trains the latent trajectory to point at the answer embedding, so the observed performance gain does not by itself demonstrate that the model has learned a general preference-reasoning process. The paper should either remove or re-purpose this loss in a leakage analysis, or show that the gain vanishes when e_tgt is replaced by a non-target embedding of similar semantic type.
  3. [Tables 3, 5, and 6] The ablation tables and the reasoning-strategy comparison report single-run numbers without standard deviations or significance tests. For example, Table 3 shows a drop from 0.1073 to 0.0948 in Toys H@10 when step-level alignment is removed, and Table 5 shows a Target-Oriented vs Non-Target CoT difference of 0.0013 in Instruments H@10; neither difference is accompanied by variance or a statistical test. Given that the leakage concern in Section 4.4.6 rests entirely on Table 5, the authors should report multiple seeds with error bars and a significance test for all ablations and for the Target vs Non-Target comparison.
  4. [Section 4.4.5, Table 4] The text states that latency is 'nearly on par with that of the non-reasoning method', but Table 4 shows Ours at 0.34s vs TALLRec at 0.23s on Instruments (about 48% higher) and 0.67s vs 0.56s on Industry (about 20% higher). The claim of near-parity should be softened or quantified, and the comparison should include a latent-reasoning baseline such as LatentR3 to show where the overhead comes from.
  5. [Section 4.5, Table 6] The online A/B test reports Exposure +0.46%, Conversion +1.39%, and Costs +2.93% and calls all of them improvements, but an increase in Costs is normally unfavorable unless 'Costs' denotes a cost-efficiency metric in the opposite direction. The paper should clarify the metric direction, report the test duration, sample sizes, and p-values, and fix the typo 'for a weak' to 'for a week'.
minor comments (6)
  1. [Section 4.1.4] The heading 'Implemenation' is misspelled; it should read 'Implementation'.
  2. [Figure 7] The x-axis label 'simga δ' is a typo for 'sigma'.
  3. [Tables 1 and 4] Dataset naming is inconsistent: the paper refers to 'MovieLens-100K' in Section 4.1.1 but the tables use 'Movie' and Table 4 uses 'MovieLens'. Please standardize the names.
  4. [Table 1] The column 'Sparsity' is not defined in the text; please state the formula (e.g., 1 - #interactions/(#users × #items)) or remove the column.
  5. [Section 3.2.1, Figure 3] The example prompt in Figure 3 places the next item under '<NextItemUserLikes>' and instructs the model that the analysis must not include information related to that item; this instruction is about the generated text, not about the hidden states, which is precisely the leakage concern. The figure should be annotated to clarify what guarantee is actually claimed.
  6. [Reproducibility] No code or data release is mentioned. Given the number of hyperparameters (alpha, gamma, sigma, lambda, K, tau1, GRPO beta/epsilon, rollout count), a reproducibility statement or code link would strengthen the paper.

Circularity Check

2 steps flagged · score 6.0 of 10

The 'latent reasoning' mechanism is partly constructed: the target item is injected into the teacher prompt and explicitly pulled into every latent update by Eq. (6), so the no-leakage claim in Sec. 4.4.6 is untested and the reasoning story reduces, in part, to answer distillation.

  1. self definitional [Section 3.2.2, Eq. (6)]
    "Building upon this, we define the actual evolutionary direction of the (k+1)th latent reasoning step as v_{k+1} = z_{k+1} - z_k. ... the ideal update vector corresponds to the direction of the target item itself, scaled by a factor of 1/(K-1). We constrain the reasoning process by maximizing the cosine consistency between the actual update vector v_{k+1} and the scaled target vector."

    The 'reasoning direction' is not discovered or measured; it is defined as cosine similarity to the target item embedding, and the loss explicitly moves every latent update toward the answer. After latent pre-training, z_K approximately equals z_1 plus the target vector e_tgt by construction, so any observed progression of latent states toward the target is the optimization objective itself, not evidence of emergent reasoning. The paper's claim that latent states 'progressively point to the final recommendation target' is therefore a restatement of the training loss rather than an independent empirical finding.

  2. fitted input called prediction [Section 3.2.1, Eqs. (3)-(4) and Section 4.4.6]
    "for each training instance (S_u, i_tgt), we feed both the user history S_u and the ground-truth target item i_tgt into a powerful reasoning model like Qwen-72B ... h_k = LLM_last(prompt_cot(S_u, i_tgt), c_{1:k}) ... we observed that Target-Oriented CoT achieves better performance, indicating that target-oriented CoT reasoning can generate higher-quality user preference trajectories without information leakage."

    The teacher states h_k are computed from a prompt whose context contains the target item; only the surface CoT text is forbidden from naming it. Step-level alignment then trains z_k to match h_k, so target-specific information can flow from the prompt into the latent states. The Section 4.4.6 conclusion that the improvement is achieved 'without information leakage' is not tested by any probe of h_k or z_k; the observed gain is exactly what target-answer distillation would produce. Thus the claim that target-oriented CoT transfers higher-quality reasoning rather than the answer is a fitted input called a prediction.

full rationale

LaRec is an empirical recommendation method, and its headline accuracy numbers are not circular in the narrow sense: they come from held-out evaluation, baseline comparisons, and ablations, and the RL stage does sample diverse paths from a user-conditioned distribution. However, the paper's mechanistic claim that latent reasoning, rather than answer leakage, drives the gains is undermined by the paper's own equations. The teacher hidden states in Eq. (3) are generated from a prompt containing the ground-truth item, so h_k can encode the answer; Eq. (4) then aligns z_k to h_k. Eq. (6) goes further and explicitly regularizes each latent update v_k toward e_tgt = MeanPooling(LLM_TKZ(i_tgt)), so after Latent Pre-training the final latent state is trained to contain the target embedding. The Section 4.4.6 assertion that Target-Oriented CoT works 'without information leakage' is not supported by any leakage probe and is contradicted by construction for the hidden states. The reported performance gains may still be real and deployable, but the specific conclusion that the model performs latent preference reasoning rather than target distillation is partly circular: the 'reasoning' trajectory is trained, by definition, to point at the target. This is partial circularity rather than full equivalence, because the final next-item prediction still requires generalization through held-out data and the RL stage adds independent signal.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The central mechanism rests on the target-conditioned teacher CoT being leakage-free, the linear accumulation assumption in Eq. 6, and the GMM user interest prior. None of these is validated independently; they define the method itself. The free parameters are training hyperparameters, most of which are tuned on validation or left unreported. No new physical or mathematical entities are postulated.

free parameters (8)
  • alpha (step-level alignment weight) = 1.0 (sensitivity: Instruments optimum 0.8, Industry 1.0)
    Weights L_SA in Eq. 7; tuned on validation and affects final performance.
  • gamma (process direction alignment weight) = 1.0
    Weights L_PA in Eq. 7; sensitivity curves show dataset-dependent optima.
  • sigma (local exploration radius) = 1.5
    Controls the Gaussian mixture width in Eq. 9; tuned via sensitivity analysis with peak at 1.5.
  • lambda (semantic reward weight) = 1.0
    Dense semantic similarity weight in Eq. 11; set manually with no full sensitivity reported.
  • K (number of latent reasoning steps) = 6
    Set to match the majority CoT length in the teacher data; later reported as the optimal value in Figure 5.
  • GRPO beta and epsilon = not reported
    KL penalty and clip range in Eq. 13 are not specified in the paper.
  • InfoNCE temperature tau1 = not reported
    Temperature in Eq. 4 is not given; it controls the sharpness of step-level contrastive alignment.
  • n (rollouts per user) = not reported
    Number of parallel latent reasoning paths sampled in Eq. 12 and Eq. 13 is not specified.
assumptions (4)
  • ad hoc to paper Linear accumulation: each latent reasoning step update v_k should form a scaled sub-component of the target item embedding e_tgt, so the ideal update is e_tgt/(K-1).
    Introduced in Eq. 6 and Section 3.2.2 to justify process direction alignment; not derived from data or theory, and it directly injects target semantics into latent states.
  • domain assumption Target-conditioned reverse CoT provides leakage-free, high-quality teacher reasoning.
    Section 3.2.1 and Figure 3: Qwen-72B generates CoT with the ground-truth item in the prompt under an instruction not to mention it; the paper assumes hidden states of this CoT do not leak target information.
  • domain assumption A uniform Gaussian mixture over historical item embeddings approximates the user's interest manifold and provides useful anchors for RL exploration.
    Eq. 9 in Section 3.3.1; there is no validation that this distribution matches the true interest manifold.
  • domain assumption Latent states and explicit CoT hidden states are comparable via cosine similarity after a lightweight reasoning head.
    Section 3.2.1 assumes representational compatibility in the shared embedding space for the InfoNCE alignment in Eq. 4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation." pith.science (2026). https://pith.science/paper/CX77ET4T

@misc{pith2026260724617,
  author       = {Pith},
  title        = {Pith review of: LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CX77ET4T}},
  note         = {Machine review of arXiv:2607.24617}
}
abstract

Large Language Models (LLMs) have shown great promise in recommendation due to superior reasoning abilities. However, existing methods mainly rely on explicit Chain-of-Thought (CoT), resulting in verbose reasoning texts and inefficient response times. latent reasoning aims to balance efficiency by thinking within a continuous latent space, yet it faces two major challenges: (1) Lack of Fine-grained Supervision: Latent reasoning relies solely on feedback from the final labels, providing sparse supervisory signals that struggle to effectively guide the optimization of multiple hidden reasoning steps. (2) Single Reasoning Path: The deterministic nature of latent reasoning impedes the exploration of users' diverse interests and preferences, thereby limiting the recommendation capabilities of LLMs. To address these issues, we propose \textbf{$LaRec$}, an efficient generative recommendation framework designed to unleash the potential of latent reasoning in LLMs. $LaRec$ consists of two core stages: First, we design Latent Pre-training that empowers LLMs with latent reasoning capabilities by providing rich supervisory signals to the latent space reasoning via step-level alignment and process direction alignment. Second, we introduce Personalized RL-tuning. Specifically, we construct a personalized Gaussian Mixture Distribution for each user based on their historical interests. By randomly sampling distinct reasoning starting points from this distribution during training, we guide the LLMs to traverse diverse reasoning paths within the latent space, enabling efficient exploration of user's multi-faceted interests. Experiments on multiple datasets show that $LaRec$ significantly outperforms existing baselines with comparable efficiency.

Figures

Figures reproduced from arXiv: 2607.24617 by the authors.

Figure 1
Figure 1. Left: Explicit Reason for Recommendation; Right: [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of 𝐿𝑎𝑅𝑒𝑐. It mainly consists of two core stages: Latent Pre-Training and Personalized RL-tuning. You are a recommendation master capable of deeply analyzing user preferences based on their historical interactions. \nThe following are the items the user has interacted with: <Historical Interactions> title:Gretsch G9500 Jim Dandy Flat To p-Vintage Sunburst, category: Musical Instruments-Guitars-Acoustic Gu… view at source ↗
Figure 3
Figure 3. Preference Reasoning Prompt: Example prompt [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Analysis of Reasoning Process Effectiveness. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Impact of Reasoning Length. 0.092 0.094 0.096 0.098 0.1 0.118 0.12 0.122 0.124 0.126 0.128 0.5 1.5 3 7 H@10 N@10 Model Size (B) Instruments H@10 N@10 0.235 0.24 0.245 0.25 0.255 0.26 0.39 0.395 0.4 0.405 0.41 0.415 0.42 0.5 1.5 3 7 H@10 N@10 Model Size (B) Industry H@1…
Figure 6
Figure 6. Figure 6: Impact of LLM Scaling Laws. 4.4.2 Impact of Reasoning Length. To analyze the impact of the number of latent reasoning steps, we vary the reasoning steps 𝐾 within the set {0, 2, 4, 6, 8} (where 0 indicates no reasoning). The experimental results in [PITH_FULL_IMAGE:fig…
Figure 7
Figure 7. Figure 7: Impact of different alignment coefficients [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 3 canonical work pages

  1. [1]

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

    arXiv 2023

  2. [2]

    Keqin Bao, Jizhi Zhang, Wenjie Wang, Yang Zhang, Zhengyi Yang, Yanchen Luo, Chong Chen, Fuli Feng, and Qi Tian. 2025. A bi-step grounding paradigm for large language models in recommendation systems.ACM Transactions on Recommender Systems3, 4 (2025), 1–27

    2025

  3. [3]

    Keqin Bao, Jizhi Zhang, Yang Zhang, Xinyue Huo, Chong Chen, and Fuli Feng

  4. [4]

    Keqin Bao, Jizhi Zhang, Yang Zhang, Wenjie Wang, Fuli Feng, and Xiangnan He. 2023. Tallrec: An effective and efficient tuning framework to align large language model with recommendation. InProceedings of the 17th ACM conference on recommender systems. 1007–1014

    2023

  5. [5]

    Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu, Yuhang Zhou, Te Gao, and Wanxiang Che. 2025. Towards reason- ing era: A survey of long chain-of-thought for reasoning large language models. arXiv preprint arXiv:2503.09567(2025)

    arXiv 2025

  6. [6]

    Zhipeng Chen, Yingqian Min, Beichen Zhang, Jie Chen, Jinhao Jiang, Daixuan Cheng, Wayne Xin Zhao, Zheng Liu, Xu Miao, Yang Lu, et al. 2025. An empir- ical study on eliciting and improving r1-like reasoning models.arXiv preprint arXiv:2503.04548(2025)

    arXiv 2025

  7. [7]

    Kairui Fu, Tao Zhang, Shuwen Xiao, Ziyang Wang, Xinming Zhang, Chenchi Zhang, Yuliang Yan, Junjun Zheng, Yu Li, Zhihong Chen, et al . 2025. Forge: Forming semantic identifiers for generative retrieval in industrial datasets.arXiv preprint arXiv:2509.20904(2025)

    arXiv 2025

  8. [8]

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

    arXiv 2024

Show all 61 references
  1. [9]

    Hao Gu, Rui Zhong, Yu Xia, Wei Yang, Chi Lu, Peng Jiang, and Kun Gai. 2025. R 4ec: A reasoning, reflection, and refinement framework for recommendation sys- tems. InProceedings of the Nineteenth ACM Conference on Recommender Systems. 411–421

    2025

  2. [10]

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al . 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948(2025)

    2025 arXiv

  3. [11]

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

    2017 arXiv

  4. [12]

    Ruidong Han, Bin Yin, Shangyu Chen, He Jiang, Fei Jiang, Xiang Li, Chi Ma, Mincong Huang, Xiaoguang Li, Chunzhen Jing, et al . 2025. Mtgr: Industrial- scale generative recommendation framework in meituan. InProceedings of the 34th ACM International Conference on Information an...

    2025

  5. [13]

    Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston, and Yuandong Tian. 2024. Training large language models to reason in a continuous latent space.arXiv preprint arXiv:2412.06769(2024)

    2024 arXiv

  6. [14]

    Yinhan He, Wendy Zheng, Yaochen Zhu, Zaiyi Zheng, Lin Su, Sriram Vasudevan, Qi Guo, Liangjie Hong, and Jundong Li. 2025. SemCoT: Accelerating Chain- of-Thought Reasoning through Semantically-Aligned Implicit Tokens.arXiv preprint arXiv:2510.24940(2025)

    2025

  7. [15]

    Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk

  8. [16]

    Chengsong Huang, Wenhao Yu, Xiaoyang Wang, Hongming Zhang, Zongxia Li, Ruosen Li, Jiaxin Huang, Haitao Mi, and Dong Yu. 2025. R-zero: Self-evolving reasoning llm from zero data.arXiv preprint arXiv:2508.05004(2025)

    2025 arXiv

  9. [17]

    Jianchao Ji, Zelong Li, Shuyuan Xu, Wenyue Hua, Yingqiang Ge, Juntao Tan, and Yongfeng Zhang. 2024. Genrec: Large language model for generative recommen- dation. InEuropean Conference on Information Retrieval. Springer, 494–502

    2024

  10. [18]

    Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recom- mendation. In2018 IEEE international conference on data mining (ICDM). IEEE, 197–206

    2018

  11. [19]

    Zhong-Zhi Li, Duzhen Zhang, Ming-Liang Zhang, Jiaxin Zhang, Zengyan Liu, Yuxuan Yao, Haotian Xu, Junhao Zheng, Pei-Jie Wang, Xiuyi Chen, et al. 2025. Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Trovato et al. From system 1 to system 2: A survey of reasoning large l...

    2025 arXiv

  12. [20]

    Enze Liu, Bowen Zheng, Xiaolei Wang, Wayne Xin Zhao, Jinpeng Wang, Sheng Chen, and Ji-Rong Wen. 2025. LARES: Latent Reasoning for Sequential Recom- mendation.arXiv preprint arXiv:2505.16865(2025)

    2025 arXiv

  13. [21]

    Zhanyu Liu, Shiyao Wang, Xingmei Wang, Rongzhou Zhang, Jiaxin Deng, Honghui Bao, Jinghao Zhang, Wuchao Li, Pengfei Zheng, Xiangyu Wu, et al

  14. [22]

    Avinash Patil and Aryan Jadon. 2025. Advancing reasoning in large language models: Promising methods and approaches.arXiv preprint arXiv:2502.03671 (2025)

    2025 arXiv

  15. [23]

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Tran, Jonah Samost, et al

  16. [24]

    Zhenyi Shen, Hanqi Yan, Linhai Zhang, Zhanghao Hu, Yali Du, and Yulan He. 2025. Codi: Compressing chain-of-thought into continuous space via self-distillation. arXiv preprint arXiv:2502.21074(2025)

    2025 arXiv

  17. [25]

    Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang

  18. [26]

    Wenhui Tan, Jiaze Li, Jianzhong Ju, Zhenbo Luo, Jian Luan, and Ruihua Song

  19. [27]

    Jiakai Tang, Xu Chen, Wen Chen, Jian Wu, Yuning Jiang, and Bo Zheng. 2026. Parallel Latent Reasoning for Sequential Recommendation.arXiv preprint arXiv:2601.03153(2026)

    2026

  20. [28]

    Jiakai Tang, Sunhao Dai, Teng Shi, Jun Xu, Xu Chen, Wen Chen, Wu Jian, and Yuning Jiang. 2025. Think Before Recommend: Unleashing the Latent Reasoning Power for Sequential Recommendation. CoRR abs/2503.22675 (2025). doi: 10. 48550.arXiv preprint ARXIV.2503.22675(2025)

    2025 arXiv

  21. [29]

    Kimi Team, Yifan Bai, Yiping Bao, Guanduo Chen, Jiahao Chen, Ningxin Chen, Ruijue Chen, Yanru Chen, Yuankun Chen, Yutian Chen, et al . 2025. Kimi k2: Open agentic intelligence.arXiv preprint arXiv:2507.20534(2025)

    2025 arXiv

  22. [30]

    Qwen Team. 2024. Qwen2.5: A Party of Foundation Models. https://qwenlm. github.io/blog/qwen2.5/

    2024

  23. [31]

    Ruoxi Wang, Rakesh Shivanna, Derek Cheng, Sagar Jain, Dong Lin, Lichan Hong, and Ed Chi. 2021. Dcn v2: Improved deep & cross network and practical lessons for web-scale learning to rank systems. InProceedings of the web conference 2021. 1785–1797

    2021

  24. [32]

    arXiv:2505.16552 [cs.CL] https://arxiv.org/abs/2505.16552

    Think Silently, Think Fast: Dynamic Latent Compression of LLM Reasoning Chains. arXiv:2505.16552 [cs.CL] https://arxiv.org/abs/2505.16552

  25. [33]

    Wenjie Wang, Xinyu Lin, Fuli Feng, Xiangnan He, and Tat-Seng Chua. 2023. Generative recommendation: Towards next-generation recommender paradigm. arXiv preprint arXiv:2304.03516(2023)

    2023 arXiv

  26. [34]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems35 (2022), 24824–24837

    2022

  27. [35]

    Xilin Wei, Xiaoran Liu, Yuhang Zang, Xiaoyi Dong, Yuhang Cao, Jiaqi Wang, Xipeng Qiu, and Dahua Lin. 2025. SIM-CoT: Supervised Implicit Chain-of- Thought.arXiv preprint arXiv:2509.20317(2025)

    2025

  28. [36]

    Bohong Wu, Mengzhao Chen, Xiang Luo, Shen Yan, Qifan Yu, Fan Xia, Tianqi Zhang, Hongrui Zhan, Zheng Zhong, Xun Zhou, et al . 2025. Parallel Loop Transformer for Efficient Test-Time Computation Scaling.arXiv preprint arXiv:2510.24824(2025)

    2025

  29. [37]

    Yu Xia, Rui Zhong, Hao Gu, Wei Yang, Chi Lu, Peng Jiang, and Kun Gai. 2025. Hierarchical tree search-based user lifelong behavior modeling on large language model. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval...

    2025

  30. [38]

    Wenjie Wang, Honghui Bao, Xinyu Lin, Jizhi Zhang, Yongqi Li, Fuli Feng, See- Kiong Ng, and Tat-Seng Chua. 2024. Learnable item tokenization for generative recommendation. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management. 2400–2409

    2024

  31. [39]

    Fengli Xu, Qianyue Hao, Zefang Zong, Jingwei Wang, Yunke Zhang, Jingyi Wang, Xiaochong Lan, Jiahui Gong, Tianjian Ouyang, Fanjin Meng, et al. 2025. Towards large reasoning models: A survey of reinforced reasoning with large language models.arXiv preprint arXiv:2501.09686(2025)

    2025 arXiv

  32. [40]

    Yi Xu, Moyu Zhang, Chenxuan Li, Zhihao Liao, Haibo Xing, Hao Deng, Jinxin Hu, Yu Zhang, Xiaoyi Zeng, and Jing Zhang. 2025. Mmq: Multimodal mixture- of-quantization tokenization for semantic id generation and user behavioral adaptation.arXiv preprint arXiv:2508.15281(2025)

    2025

  33. [41]

    Mengjiao Yang, Dale Schuurmans, Pieter Abbeel, and Ofir Nachum. 2022. Chain of Thought Imitation with Procedure Cloning. arXiv:2205.10816 [cs.LG] https: //arxiv.org/abs/2205.10816

    2022 arXiv

  34. [42]

    Wei Yang, Rui Zhong, Yiqun Chen, Shixuan Li, Heng Ping, Chi Lu, and Peng Jiang. 2025. FITMM: Adaptive Frequency-Aware Multimodal Recommendation via Information-Theoretic Representation Learning. InProceedings of the 33rd ACM International Conference on Multimedia. 6193–6202

    2025

  35. [43]

    Wei Yang, Rui Zhong, Yiqun Chen, Chi Lu, and Peng Jiang. 2025. Structured Spectral Reasoning for Frequency-Adaptive Multimodal Recommendation.arXiv preprint arXiv:2512.01372(2025)

    2025

  36. [44]

    Yu Xia, Rui Zhong, Zeyu Song, Wei Yang, Junchen Wan, Qingpeng Cai, Chi Lu, and Peng Jiang. 2025. Trackrec: Iterative alternating feedback with chain-of-thought via preference alignment for recommendation.arXiv preprint arXiv:2508.15388 (2025)

    2025 arXiv

  37. [45]

    Qihang Yu, Kairui Fu, Shengyu Zhang, Zheqi Lv, Fan Wu, and Fei Wu

  38. [46]

    Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhaojie Gong, Fangda Gu, Jiayuan He, et al . 2024. Actions speak louder than words: trillion-parameter sequential transducers for generative recommendations. InProceedings of the 41st International Conf...

    2024

  39. [47]

    Yang Zhang, Wenxin Xu, Xiaoyan Zhao, Wenjie Wang, Fuli Feng, Xiangnan He, and Tat-Seng Chua. 2025. Reinforced Latent Reasoning for LLM-based Recommendation.arXiv preprint arXiv:2505.19092(2025)

    2025

  40. [48]

    Rui Zhao, Rui Zhong, Haoran Zheng, Wei Yang, Chi Lu, Beihong Jin, Peng Jiang, and Kun Gai. 2025. Hierarchical Sequence ID Representation of Large Language Models for Large-scale Recommendation Systems. InCompanion Proceedings of the ACM on Web Conference 2025. 641–650

    2025

  41. [49]

    Bowen Zheng, Yupeng Hou, Hongyu Lu, Yu Chen, Wayne Xin Zhao, Ming Chen, and Ji-Rong Wen. 2024. Adapting large language models by integrating collaborative semantics for recommendation. In2024 IEEE 40th International Conference on Data Engineering (ICDE). IEEE, 1435–1448

    2024

  42. [50]

    Wencai Ye, Mingjie Sun, Shaoyun Shi, Peng Wang, Wenjin Wu, and Peng Jiang

  43. [51]

    InProceedings of the 34th ACM International Conference on Information and Knowledge Management

    DAS: Dual-Aligned Semantic IDs Empowered Industrial Recommender System. InProceedings of the 34th ACM International Conference on Information and Knowledge Management. 6217–6224

  44. [52]

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

    2018

  45. [53]

    ThinkRec: Thinking-based recommendation via LLM.arXiv preprint arXiv:2505.15091(2025)

    2025

  46. [58]

    Guorui Zhou, Jiaxin Deng, Jinghao Zhang, Kuo Cai, Lejian Ren, Qiang Luo, Qian- qian Wang, Qigen Hu, Rui Huang, Shiyao Wang, et al. 2025. OneRec Technical Report.arXiv preprint arXiv:2506.13695(2025)

    2025

  47. [59]

    Guorui Zhou, Hengrui Hu, Hongtao Cheng, Huanjie Wang, Jiaxin Deng, Jinghao Zhang, Kuo Cai, Lejian Ren, Lu Ren, Liao Yu, et al. 2025. Onerec-v2 technical report.arXiv preprint arXiv:2508.20900(2025)

    2025 arXiv

  48. [61]

    Jie Zhu, Zhifang Fan, Xiaoxie Zhu, Yuchen Jiang, Hangyu Wang, Xintian Han, Haoran Ding, Xinmin Wang, Wenlin Zhao, Zhen Gong, et al. 2025. Rankmixer: Scaling up ranking models in industrial recommenders. InProceedings of the 34th ACM International Conference on Information and ...

    2025

  49. [2015]

    Session-based recommendations with recurrent neural networks.arXiv preprint arXiv:1511.06939(2015)

    2015 arXiv

  50. [2019]

    InProceedings of the 28th ACM international conference on information and knowledge management

    BERT4Rec: Sequential recommendation with bidirectional encoder rep- resentations from transformer. InProceedings of the 28th ACM international conference on information and knowledge management. 1441–1450

  51. [2023]

    Recommender systems with generative retrieval.Advances in Neural Information Processing Systems36 (2023), 10299–10315

    2023

  52. [2024]

    InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing

    Decoding matters: Addressing amplification bias and homogeneity issue in recommendations for large language models. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 10540–10552

    2024

  53. [2025]

    Onerec-think: In-text reasoning for generative recommendation.arXiv preprint arXiv:2510.11639(2025)

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.