Pith. sign in

REVIEW 4 major objections 5 minor 28 references

Hypothesis-Driven Shelf Generation for Personalised Recommendation

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper argues that recommendation rows can be generated from per-user natural-language 'shelf hypotheses' instead of hand-written templates, and that these generated shelves stay competitive with existing ones in live traffic.

desk verdict A genuinely novel production architecture with a solid offline pipeline; the online 'competitive' claim is oversold by best-of-family selection, but the system is real and worth refereeing. read the letter →

arxiv 2607.25823 v1 pith:O4SR7E7Q submitted 2026-07-28 cs.IR

Aleksandr V. Petrov , Tarun Chillara , Matthew D. Moellman , Lucas de Haas , Yabai Song , Alina Susoykina , Melissa Crawford , Gabriel Negash
show 8 more authors
Erik Franco Tasnim Rahman Binal Jhaveri Shubham Bansal Hugues Bouchard Roberto Mirizzi Mounia Lalmas Aloïs Gruson
This is my paper · ORCID
classification cs.IR
keywords shelfgenerationhypothesis-drivenrecommendationgenerativeretrievalSemanticIDsLLM-as-a-judgepersonalisedshelvesmulti-listmusicstreaming
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 tries to show that the themed rows on a music-streaming home page—shelves like 'More of What You Like'—need not come from a fixed set of hand-authored templates. Instead, the system can generate a natural-language hypothesis for each user about what a personalised shelf should contain, then retrieve catalogue items that instantiate that hypothesis, and finally align the shelf's title and contents. The authors claim this decoupling of planning from retrieval substantially expands the supply of personalised shelves while keeping engagement competitive with strong existing shelves in several content types. A sympathetic reader would care because it points to a scalable way to serve long-tail tastes without manually maintaining a template for every niche.

What carries the argument

The load-bearing object is the 'shelf hypothesis', a structured tuple h=(q,c,f,r,t0,d0): a natural-language shelf description q, a target content type c, a familiarity level f, optional routing constraints r, and provisional title and subtitle. It functions as a compact contract between the planning stage and the fulfilment stage, letting each be optimised and evaluated independently. Fulfilment is performed by constrained generative retrieval: a small language model generates Semantic IDs (compact discrete identifiers for catalogue entities) decoded through content-type-specific tries, so every generated identifier resolves to a valid album, artist, playlist, podcast show, or episode.

What would settle it

Collect human quality ratings on a random sample of the 800 user hypotheses and 10,000 fulfilled shelves, and compare them to the LLM judge scores; low correlation would undermine the offline-stage claims. Alternatively, run the generated shelves under the standard production ranking policy rather than uniform random exposure and test whether the engagement advantage persists when position and ranker selection are controlled.

Watch

Extended reading notes

Core claim

The central claim is that shelf recommendation can be reframed as a generative planning problem: produce a structured natural-language 'shelf hypothesis' from a user's behaviour, treat that hypothesis as an intermediate contract between planning and catalogue retrieval, and then fulfil it with constrained generative retrieval over Semantic IDs. The paper reports that this hypothesis-driven pipeline, running fully offline with precomputed serving, produces shelves that under uniform random exposure on the home page achieve the strongest engagement in album and episode pools, rank second in artist and show pools, and lag in playlist and show pools—while greatly expanding the diversity of perso

Load-bearing premise

The offline conclusions for hypothesis quality, generative-retrieval superiority, and alignment gains rest on in-house LLM-as-a-judge rubrics whose scores are assumed to track human judgement, and no task-specific human-alignment or inter-rater validation is reported.

Editorial extensions

If this is right

  • If the central claim is correct, home-page shelves can be generated per user rather than selected from a hand-maintained template inventory, enabling long-tail niches such as 'glacial ambient post-rock with orchestral textures'.
  • The four-stage decomposition (hypothesis, fulfilment, alignment, serving) allows each stage to be swapped or ablated independently, so retrieval quality and presentation quality can be optimised and measured separately.
  • Generative retrieval over Semantic IDs appears to capture catalogue associations that lexical and embedding-based baselines miss, with the largest judged gains in completeness, diversity, and hypothesis coverage.
  • The alignment stage, which selects the final items and rewrites the title to match them, produces the largest quality jump in the pipeline, especially in title-promise fulfilment (+99% under the judge rubric).
  • In online evaluation under uniform random exposure, generated shelves beat the strongest existing shelves in album and episode pools and remain competitive in several other pools, though podcast shows remain a clear weakness.

Reading between the lines

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

  • The 'hypothesis as contract' pattern is domain-agnostic: any multi-list surface (news, video, e-commerce) with implicit row promises could use the same separation between planning and fulfilment, with the hypothesis schema adapted to local entities.
  • If LLM-as-a-judge scores are validated against human raters for this open-ended task, the offline pipeline could become a fast, low-cost development loop for shelf generation, reducing the need for repeated online experiments.
  • The large gap in the podcast show pool suggests that spoken-word shelves may need a different hypothesis schema, a different fulfilment index, or content-type-specific alignment rules; this is directly testable by isolating show-pool generation and tuning it.
  • Because planning and fulfilment run fully offline, the authors' stated future direction of near-real-time shelf generation is plausible: only the final serving lookup would need to be online, so latency constraints may not block faster adaptation to evolving user behaviour.
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 describes a production system at Spotify that generates personalised Home shelves from natural-language "shelf hypotheses" rather than from a fixed inventory of hand-authored templates. The architecture decomposes the pipeline into four stages: hypothesis generation (with distillation from a frontier LLM), constrained generative retrieval over Semantic IDs, candidate selection and shelf alignment, and fully offline serving. The evaluation mixes offline LLM-as-a-judge analyses of hypothesis quality, fulfilment, and alignment with an online study under uniform random exposure on Home. The central claim is that hypothesis-driven shelves substantially expand personalised recommendation supply and achieve engagement that is competitive with strong existing shelves in some content types. The architectural decomposition and the online randomised-exposure protocol are genuine strengths, but I find that the offline conclusions rest on an unvalidated judge, the online 'competitive' claim is supported only by a best-of-family comparison, and the supply-expansion claim is not operationalised.

Significance. If the claims were fully supported, this would be a notable industrial contribution: it shows how to replace fixed shelf templates with generated natural-language hypotheses as intermediate planning representations, how to separate planning from retrieval, and how to ground those hypotheses in catalogue entities via constrained generative retrieval over Semantic IDs. The paper is honest in reporting confidence intervals, Bonferroni corrections, and the descriptive nature of the online comparisons, and the uniform-random-exposure online protocol is a meaningful, LLM-independent behavioural signal. The main scientific value is the architectural decomposition and the stage-specific evaluation hooks, which could be reused by other production systems. However, the evidence currently falls short of the headline claims: the offline results are only as credible as the LLM judge, the online 'competitive' conclusion is based on selected maxima rather than family-level comparisons, and 'substantially expand supply' is never measured.

major comments (4)
  1. [Section 4.1, Tables 3, 5, 6] All offline conclusions—hypothesis quality, Generative Retrieval beating BM25/MiniLM, and the +78% alignment gain—are measured by in-house LLM judges on a 0–2 ordinal scale. No task-specific human calibration or inter-rater reliability is reported. The cited evidence [14, 26] is for general or Cranfield-style relevance judgements, not for open-ended shelf-hypothesis and shelf-set quality, where the rubrics involve subjective constructs such as 'Discovery Potential' and 'Title Promise Fulfilment'. The paper itself calls these 'directional offline signals', but the abstract and Section 4 use them to support concrete architectural claims. Without a human-alignment sample or at least a strong robustness analysis (e.g., agreement statistics on this task), Tables 3, 5, and 6 cannot be taken as evidence for the offline claims.
  2. [Section 4.5, Table 7] The abstract's 'competitive with strong existing shelves in some settings' is supported almost entirely by Table 7, which reports only the strongest hypothesis-driven shelf and the strongest classic shelf observed in each content-type pool. The number of candidates per pool is not given, so selecting the maximum over an unknown-sized family of generated shelves can inflate the apparent advantage through winner's curse. The album +36% and episode +2% deltas are therefore not evidence that the hypothesis-driven shelf family is competitive; they are evidence only that at least one member performed well in that pool. The table's own caption warns that these are 'descriptive within-pool comparisons, not pooled causal effect estimates', but the narrative in Section 4.5 and the abstract generalises beyond this. Please report the full distribution or family-level means, the number of candidates
  3. [Abstract, Section 4.5, Section 5] The headline claim that hypothesis-driven shelves 'substantially expand personalised recommendation supply' is never operationalised. No counts are given of generated hypotheses per user, fulfilled shelves per user, served candidate shelves, or any measure of how supply is expanded relative to the template inventory. The only evidence is architectural (additional ranking candidates) plus the qualitative statement in Section 5. Since this is one of the two headline contributions, it needs a concrete metric—for example, number/coverage of generated shelf types per user, or the increase in eligible shelf candidates entering Home ranking—or it should be removed from the abstract and conclusions.
  4. [Section 4.4, Table 6] The pre/post alignment comparison is between two disjoint cohorts (10,000 pre-alignment shelves vs. 16,000 post-alignment shelves with 'no shared shelf identifiers'). This is not a paired before/after evaluation, so the +78% overall improvement and the per-dimension gains could reflect differences in the underlying hypotheses or users rather than the effect of Stage 3. To isolate the alignment stage, the same fulfilled shelves should be evaluated before and after alignment, or at least the cohorts should be shown to be comparable on hypothesis-quality and content-type distributions. As written, the strong causal language ('candidate selection and shelf alignment substantially improve') is not supported by the design.
minor comments (5)
  1. [Section 3.2] The text refers to a 'compact open-source LLM' and a 'distilled open-source generator', but no model identifier, size, or version is given. For reproducibility and for readers to understand the production footprint, please name the base model and the distillation procedure.
  2. [Section 4.3, Table 3] The caption reports paired t-tests over '10k resamples', but the unit of analysis is unclear: is the pairing over individual shelf hypotheses? Please state the number of distinct hypotheses and users, and whether the bootstrap resamples are at the hypothesis or user level.
  3. [Section 4.5, Table 7] The table mentions 'content-type shuffle pools' but does not define what a shuffle pool is, how many pools per content type were analysed, or how the 'best classic' shelf was selected. A brief definition and the number of pools would make the description interpretable.
  4. [Section 4.5] The text says 'a small fraction of Home requests are assigned' to uniform random exploration, but the fraction is not quantified. Reporting the exploration fraction and request counts would help readers assess the precision of Table 7.
  5. [Section 4.2] The distillation parity check on 800 user profiles uses the same LLM judge as the generator comparison. This is a useful internal check, but without human labels it does not establish that the distilled model preserves quality in an absolute sense. Consider adding a small human-evaluated sample to anchor the judge.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims rest on online behavioral data and explicitly directional offline signals, not on by-construction reductions.

full rationale

The paper does not derive a target result from a fitted parameter or from a self-referential definition. The offline LLM-as-a-judge comparisons (GR vs baselines in Table 3, alignment gains in Table 6) are the only plausibly self-referential components: the rubrics are designed in-house and no task-specific human calibration is reported. But the paper explicitly frames these as 'directional offline signals for optimisation and failure-mode analysis' and treats 'online user behaviour as the final measure of shelf quality' (Section 4.1). No parameter learned from the LLM judge is then used as an input to the online claim, so there is no fitted-input-called-prediction structure. The judge-validity citation [14] overlaps with the present authors, but it is an independent empirical study and is accompanied by an external citation [26] on LLM-judge biases; the paper also discloses that judges have biases that require careful protocol design. The Table 7 online comparison does select the strongest observed shelf per pool, which creates a winner's-curse/statistical-inference concern and the paper itself cautions that these are 'descriptive within-pool comparisons, not pooled causal effect estimates.' That is a validity limitation, not a circular reduction: the comparison is not identical to its inputs by construction. The 'substantially expand supply' claim is not operationalised with counts, so it is unsupported, but it is an architectural consequence of adding generated candidates rather than a predicted quantity fitted to the data that reports it. Overall, the central online engagement claim is measured by user behavior under randomized exposure and does not reduce to the system's own definitions or to a self-citation chain.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The paper contains no mathematical derivation; its claims rest on evaluation-instrument assumptions (LLM judges, randomized exposure), representational assumptions (the shelf hypothesis as a sufficient contract), and grounding assumptions (SemID generative retrieval). None are fitted free parameters in the classical sense, but several are chosen by hand and undisclosed.

free parameters (4)
  • Hybrid baseline weight alpha = 0.5
    BM25 and MiniLM scores are linearly interpolated with equal weights for the Hybrid baseline in Table 3; this choice is by hand, but it is a baseline rather than part of the proposed system.
  • LLM judge ordinal scale = 0-2
    The authors chose a 0-2 coarse ordinal scale for both judges to balance discrimination and stability; all offline scores and significance tests depend on this choice.
  • Uniform random exploration fraction = not disclosed ('a small fraction')
    The online evaluation uses an undisclosed fraction of Home requests; this determines sample size and precision of the engagement estimates.
  • Final shelf size k = not disclosed
    Stage 3 selects final k items per shelf; k affects coherence, title-promise, and online engagement but is not reported.
assumptions (5)
  • ad hoc to paper LLM-as-a-judge scores approximate human judgements for shelf-generation quality without task-specific calibration.
    The paper relies on prior evidence [14,26] that LLM judges track human judgements, but does not validate the rubrics against human labels or report inter-rater agreement in this setting.
  • domain assumption Uniform random exposure provides an unbiased comparison of shelf quality independent of ranker/position effects.
    Engagement under randomized ordering is assumed to isolate shelf quality from position/ranker bias; the paper treats results as descriptive, not causal.
  • domain assumption Semantic IDs and constrained trie decoding provide valid catalogue grounding for generated hypotheses.
    The fulfilment model presupposes that SemIDs are learnable, resolvable identifiers and that content-type tries guarantee entity validity; this is based on cited prior work [5,9,16], not on evidence in this paper.
  • ad hoc to paper A natural-language shelf hypothesis is a sufficient intermediate planning representation to capture long-tail taste and constrain retrieval.
    This is the paper's core architectural postulate; no independent benchmark establishes the expressiveness/sufficiency of the hypothesis format.
  • domain assumption Distilled-model parity measured on 800 user profiles generalizes to production across all users.
    Parity check uses a fixed cohort from a single snapshot; generalization to all production profiles is assumed.
invented entities (1)
  • Shelf hypothesis h=(q,c,f,r,t0,d0)
    purpose: Intermediate representation between user modelling and catalogue retrieval; decouples shelf planning from fulfilment.
    Central construct of the paper; a proposed data structure whose sufficiency as a planning/retrieval contract is assumed, not independently evidenced.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hypothesis-Driven Shelf Generation for Personalised Recommendation." pith.science (2026). https://pith.science/paper/O4SR7E7Q

@misc{pith2026260725823,
  author       = {Pith},
  title        = {Pith review of: Hypothesis-Driven Shelf Generation for Personalised Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/O4SR7E7Q}},
  note         = {Machine review of arXiv:2607.25823}
}
read the original abstract

Modern recommendation interfaces organise content into shelves: themed rows such as "More of What You Like" or "New Releases for You." In production systems, these shelves are typically defined through hand-crafted templates coupled with dedicated retrieval logic. While effective for broad recommendation intents, this approach does not scale to the long tail of individual taste. We present a content-hypothesis-driven shelf generation system for Spotify Home that replaces fixed templates with natural-language hypotheses describing what a personalised shelf should contain. The system has four stages hypothesis generation, catalogue fulfilment, shelf alignment, and offline serving. This decomposition decouples shelf planning from catalogue fulfilment, supports independent optimisation of planning and retrieval, and enables both constrained generative retrieval over catalogue entities and distillation of frontier LLM behaviour into compact models. Our production pipeline combines hypothesis generation, generative retrieval, candidate selection and shelf alignment, offline LLM-as-a-judge evaluation, and precomputed serving. We describe the end-to-end architecture and evaluate it through offline analyses and an early online evaluation under uniform random exposure on Spotify Home. Results show that hypothesis-driven shelves substantially expand personalised recommendation supply with engagement that varies by content type and is competitive with strong existing shelves in some settings.

Figures

Figures reproduced from arXiv: 2607.25823 by the authors.

Figure 1
Figure 1. Examples of hypothesis-driven personalised shelves [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Conceptual pipeline. User-profile information is consumed only by hypothesis generation. Downstream fulfilment, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Hypothesis-to-Shelf Judge score by content type [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 2 canonical work pages

  1. [1]

    Walid Bendada, Guillaume Salha, and Théo Bontempelli. 2020. Carousel Per- sonalization in Music Streaming Apps with Contextual Bandits. InRecSys 2020: Fourteenth ACM Conference on Recommender Systems, Virtual Event, Brazil, Sep- tember 22–26, 2020. ACM, 420–425. doi:10.1145/3383313.3412217

    arXiv 2020

  2. [3]

    Deborah Cohen, Michal Aharon, Yair Koren, Oren Somekh, and Raz Nissim

  3. [4]

    Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep Neural Networks for YouTube Recommendations. InProceedings of the 10th ACM Conference on Recommender Systems, RecSys 2016, Boston, MA, USA, September 15–19, 2016. ACM, 191–198. doi:10.1145/2959100.2959190

    arXiv 2016

  4. [5]

    Edoardo D’Amico, Marco De Nadai, Praveen Chandar, Divita Vohra, Shawn Lin, Max Lefarov, Paul Gigioli, Gustavo Penha, Ilya Kopysitsky, Ivo Joel Senese, et al

  5. [6]

    Mathieu Delcluze, Antoine Khoury, Clémence Vast, Valerio Arnaudo, Léa Briand, Walid Bendada, and Thomas Bouabça. 2025. Text2Playlist: Generating Person- alized Playlists from Text on Deezer. InAdvances in Information Retrieval: 47th European Conference on Information Retrieval, ECIR 2025 (Lecture Notes in Com- puter Science, Vol. 15576). Springer, 164–170...

    work page doi:10.1007/978-3-031-88720-8_27 2025

  6. [7]

    Shijie Geng, Shuchang Liu, Zuohui Fu, Yingqiang Ge, and Yongfeng Zhang. 2022. Recommendation as Language Processing (RLP): A Unified Pretrain, Personalized Prompt & Predict Paradigm (P5). InProceedings of the 16th ACM Conference on Recommender Systems (RecSys ’22). ACM, 299–315. doi:10.1145/3523227.3546767

    arXiv 2022

  7. [8]

    Carlos Alberto Gomez-Uribe and Neil Hunt. 2016. The Netflix Recommender System: Algorithms, Business Value, and Innovation.ACM Trans. Manag. Inf. Syst.6, 4 (2016), 13:1–13:19. doi:10.1145/2843948

    doi:10.1145/2843948 2016

  8. [9]

    Ruining He, Lukasz Heldt, Lichan Hong, Raghunandan Keshavan, Shifan Mao, Nikhil Mehta, Zhengyang Su, Alicia Tsai, Yueqi Wang, Shao-Chuan Wang, et al

Show all 28 references
  1. [10]

    Benedikt Loepp. 2023. Multi-list interfaces for recommender systems: survey and future directions.Frontiers in Big Data6 (2023), 1239705. doi:10.3389/fdata. 2023.1239705

    2023

  2. [11]

    Benedikt Loepp and Jürgen Ziegler. 2023. How Users Ride the Carousel: Exploring the Design of Multi-List Recommender Interfaces From a User Perspective. In Proceedings of the 17th ACM Conference on Recommender Systems, RecSys 2023, Singapore, September 18–22, 2023. ACM, 1090–1...

    2023

  3. [12]

    InProceedings of the ACM Web Conference 2026

    Plum: Adapting pre-trained language models for industrial-scale generative recommendations. InProceedings of the ACM Web Conference 2026. 8093–8104

    2026

  4. [13]

    Gustavo Penha, Edoardo D’Amico, Marco De Nadai, Enrico Palumbo, Alexandre Tamborrino, Ali Vardasbi, Max Lefarov, Shawn Lin, Timothy Heath, Francesco Hypothesis-Driven Shelf Generation for Personalised Recommendation RecSys ’26, September 27-October 02, 2026, Minneapolis, MN, U...

    2026

  5. [14]

    Petrov, Claudia Hauff, Enrico Palumbo, Ali Var- dasbi, Edoardo D’Amico, Francesco Fabbri, Alice Wang, Praveen Chandar, Henrik Lindström, Hugues Bouchard, and Mounia Lalmas

    Gustavo Penha, Aleksandr V. Petrov, Claudia Hauff, Enrico Palumbo, Ali Var- dasbi, Edoardo D’Amico, Francesco Fabbri, Alice Wang, Praveen Chandar, Henrik Lindström, Hugues Bouchard, and Mounia Lalmas. 2025. Do LLM-judges Align with Human Relevance in Cranfield-style Recommende...

    2025 doi

  6. [15]

    Enrico Palumbo, Gustavo Penha, Andreas Damianou, José Luis Redondo García, Timothy Christopher Heath, Alice Wang, Hugues Bouchard, and Mounia Lal- mas. 2025. Text2Tracks: Prompt-based Music Recommendation via Generative Retrieval.arXiv preprint arXiv:2503.24193(2025). arXiv:25...

    2025 arXiv

  7. [16]

    Tran, Jonah Samost, Maciej Kula, Ed H

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Q. Tran, Jonah Samost, Maciej Kula, Ed H. Chi, and Maheswaran Sathiamoorthy. 2023. Recommender Systems with Generative Retrieval. InAdvances in Neural In...

    2023

  8. [17]

    Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford

    Stephen E. Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford. 1994. Okapi at TREC-3. InProceedings of The Third Text REtrieval Conference, TREC 1994, Gaithersburg, Maryland, USA, November 2-4, 1994 (NIST Special Publication), Donna K. Harman (E...

    1994

  9. [18]

    Behnam Rahdari, Branislav Kveton, and Peter Brusilovsky. 2022. The Magic of Carousels: Single vs. Multi-List Recommender Systems. InHT ’22: 33rd ACM Conference on Hypertext and Social Media, Barcelona, Spain, June 28 – July 1, 2022. ACM, 166–174. doi:10.1145/3511095.3531278

    2022

  10. [19]

    Hendryx, Brad Kenstler, and Bing Liu

    Manasi Sharma, Chen Bo Calvin Zhang, Chaithanya Bandi, Clinton Wang, Ankit Aich, Huy Nghiem, Tahseen Rabbani, Ye Htet, Brian Jang, Sumana Basu, Aish- warya Balwani, Denis Peskoff, Marcos Ayestaran, Sean M. Hendryx, Brad Kenstler, and Bing Liu. 2026. ResearchRubrics: A Benchmar...

    2026

  11. [20]

    Sanidhya Singal, Piyush Singh, and Manjeet Dahiya. 2021. Automatic Collection Creation and Recommendation. InRecSys ’21: Fifteenth ACM Conference on Rec- ommender Systems, Amsterdam, The Netherlands, September 27 – October 1, 2021. ACM, 633–638. doi:10.1145/3460231.3478865

    2021

  12. [21]

    Sentence Transformers. 2021. sentence-transformers/all-MiniLM-L12-v2. https: //huggingface.co/sentence-transformers/all-MiniLM-L12-v2. Hugging Face model card

    2021

  13. [22]

    Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Gupta, Tal Schuster, William W

    Yi Tay, Vinh Q. Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Gupta, Tal Schuster, William W. Cohen, and Donald Metzler. 2022. Transformer Memory as a Differentiable Search Index. In Advances in Neural Information Processing Syste...

    2022

  14. [23]

    Federico Tomasi, Francesco Fabbri, Justin Carter, Elias Kalomiris, Mounia Lal- mas, and Zhenwen Dai. 2025. Prompt-to-Slate: Diffusion Models for Prompt- Conditioned Slate Generation. InProceedings of the 19th ACM Conference on Recommender Systems (RecSys ’25). ACM. doi:10.1145...

    2025

  15. [24]

    Chi, and Xinyang Yi

    Anima Singh, Trung Vu, Nikhil Mehta, Raghunandan Keshavan, Maheswaran Sathiamoorthy, Yilin Zheng, Lichan Hong, Lukasz Heldt, Li Wei, Devansh Tandon, Ed H. Chi, and Xinyang Yi. 2024. Better Generalization with Semantic IDs: A Case Study in Ranking for Recommendations. InProceed...

    2024

  16. [25]

    Yujing Wang, Yingyan Hou, Haonan Wang, Ziming Miao, Shibin Wu, Qi Chen, Yuqing Xia, Chengmin Chi, Guoshuai Zhao, Zheng Liu, Xing Xie, Hao Sun, Weiwei Deng, Qi Zhang, and Mao Yang. 2022. A Neural Corpus Indexer for Document Retrieval. InAdvances in Neural Information Processing...

    2022

  17. [26]

    Xing, Hao Zhang, Joseph E

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena. InAdvances in Neural Information Pr...

    2023

  18. [27]

    Victor Wang, Michael J. Q. Zhang, and Eunsol Choi. 2025. Improving LLM-as-a- Judge Inference with the Judgment Distribution. InFindings of the Association for Computational Linguistics: EMNLP 2025. Association for Computational Lin- guistics, Suzhou, China, 23173–23199. doi:10...

    2025 doi

  19. [2017]

    Expediting Exploration by Attribute-to-Feature Mapping for Cold-Start Recommendations. InProceedings of the Eleventh ACM Conference on Recom- mender Systems, RecSys 2017, Como, Italy, August 27-31, 2017, Paolo Cremonesi, Francesco Ricci, Shlomo Berkovsky, and Alexander Tuzhili...

    2017

  20. [2026]

    Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify.arXiv preprint arXiv:2603.17540(2026)

    2026

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.