Pith. sign in

REVIEW 4 major objections 6 minor 2 cited by

Revealing Potential Biases in LLM-Based Recommender Systems in the Cold Start Setting

T0 review · 4 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Cold-start LLM recommenders condition their top-20 picks on sensitive attributes in ways that reproduce gendered and Western-cultural stereotypes, and model scale does not fix this bias.

desk verdict Worth reading as a benchmark proposal with one solid gender-stereotype result, but the cultural-bias pillar lacks a catalog base-rate control and the headline claims outrun the evidence. read the letter →

arxiv 2508.20401 v2 pith:TIZQXQQJ submitted 2025-08-28 cs.IR cs.LG

Alexandre Andre , Gauthier Roy , Eva Dyer , Kai Wang This is my paper · ORCID
classification cs.IRcs.LG
keywords LanguagemodelsRecommendersystemsColdstartFairnessBiasbenchmarkGenderstereotypesCulturalModelscaling
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

This paper argues that when a recommendation system has almost no information about a new user—only attributes such as gender, age, or language—LLM-based recommenders do not behave neutrally: they systematically adjust their suggestions to those attributes, reproducing gendered and cultural stereotypes. To demonstrate this, it builds a benchmark in which an LLM must re-rank a fixed 500-item catalog into a top-20 list for a neutral user or for a user described by a single sensitive attribute, and measures how far the attribute-conditioned lists diverge from the neutral one. Applying the benchmark to Gemma 3 and Llama 3.2 models across music, movies, and colleges, the paper reports consistent bias across all three domains, including a strong default preference for Western content (91.3% of the 'neutral' user's movie list is Western) and a non-linear relation between model size and divergence, with the 4B model diverging less than the 1B or 12B variants. The paper offers the benchmark as a reusable audit for any open-weight LLM, arguing that fairness in cold-start recommendation cannot be assumed to follow from scale or from an ostensibly attribute-free prompt.

What carries the argument

The load-bearing instrument is a counterfactual re-ranking audit pipeline. For each sensitive attribute (gender, nationality, religion, parent's occupation), the pipeline builds a prompt that asks the LLM to pick and order the top 20 items from a fixed 500-item catalog, alongside a neutral prompt that names no attribute; the difference between the two lists is the bias signal. The similarity measures—Jaccard/IOU (set overlap), SERP (rank-weighted overlap), and PRAG (pairwise rank-order agreement)—are each converted to a divergence score B(a)=1−Sim, so 'bias' is operationalized as the distance from neutral. The pipeline runs on any open-weight model through an efficient inference backend, whi

What would settle it

Replace the neutral reference list with a random sample of the same 500-item catalog and recompute every divergence score: if the gender and Western-preference patterns disappear or reverse, the reported bias is an artifact of the baseline's composition. Separately, re-run the 1B/4B/12B scale comparison with at least 20 seeds per model and dataset; if the 4B model is no longer the least divergent on Music and Movie, the non-linear scale-fairness claim falls.

Watch

Extended reading notes

Core claim

The paper's central claim is that in zero-context (cold-start) recommendation, LLMs condition their ranked output on whatever sensitive attribute appears in the prompt, and that this conditioning tracks societal stereotypes rather than user taste. Concretely, on the movie domain, Gemma 3 4B gives the 'girl' and 'female' personas 14.8% and 18.0% action films versus 40.5% for 'a boy' and 35.0% for a neutral user; on the culture axis, the neutral, 'a Buddhist', and 'a Muslim' personas all receive 85–93% Western movies, while a Chinese or Japanese persona receives 48% and 22% respectively. The paper further claims that scale does not behave monotonically: across Gemma 1B, 4B, and 12B, the 4B mod

Load-bearing premise

The bias scores assume that a prompt naming no sensitive attribute is an unbiased reference, yet the paper's own data show that 'neutral' prompt producing 91.3% Western movies, so that reference already embodies a cultural default; if that baseline is biased, low divergence from it cannot prove low bias.

Editorial extensions

If this is right

  • If the benchmark's results generalize, platforms cannot treat larger models as an automatic fairness fix; model scale must be swept and evaluated per domain because the 4B model is not consistently the fairest either.
  • A single explicit, task-relevant preference (e.g., 'action movie fan') can overpower sensitive-attribute stereotypes in prompts, giving a cheap mitigation lever for cold-start user experiences.
  • The 'neutral' prompt is not a culture-free default: the neutral user's 91.3% Western movie share means any bias score computed against neutral inherits that Western skew, so neutral baselines need to be reported and audited themselves.
  • Adding a consequential domain—college recommendation—extends fairness audits beyond media and shopping to education and career pathways, where biased suggestions have direct life effects.

Reading between the lines

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

  • Going beyond the paper: because the neutral reference itself is Western-skewed, the divergence scores likely understate bias against non-Western personas and overstate it for Western ones; recomputing B(a) against a reference list that matches the catalog's true cultural composition would be a direct test.
  • Going beyond the paper: the paper's movie culture results suggest a corpus-language confound; prompting the same models in non-English languages, as the paper floats for future work, could separate 'Western bias' from 'English-corpus bias'.
  • Going beyond the paper: the re-ranking task limits the audit to top-k selection from a fixed catalog; a retrieval-style evaluation with live or larger catalogs could reveal whether the biases intensify or wash out when the model does not have the items already listed.
  • Going beyond the paper: the non-monotonic scale result is consistent with the 12B model over-following the attribute rather than failing at the task; a prompt-level analysis of attribute mentions in generated text would separate capability limits from stereotype amplification.
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 / 6 minor

Summary. The paper introduces a benchmark pipeline for auditing fairness of open-source LLM-based recommender systems in a cold-start setting. The pipeline takes a catalog of N items and a list of sensitive attributes, generates neutral and attribute-conditioned prompts, asks the LLM to select a top-k ranked list, and computes bias as the complement of a similarity score (IOU, SERP, PRAG) between attribute-conditioned and neutral recommendation lists. Three domains are used: movies, music, and a newly introduced college domain. Experiments on Gemma 3 (1B/4B/12B) and Llama 3.2 (3B) are reported. Four hypotheses are examined: H1 that larger models are less biased, H2 that LLMs reproduce gender stereotypes, H3 that adding user context mitigates bias, and H4 that LLMs default to Western content. The paper claims consistent biases across domains, including gendered and cultural stereotypes, and a non-linear relationship between model size and fairness.

Significance. If the claims hold, the benchmark would be a useful, reproducible tool for a practically important problem: auditing cold-start LLM recommenders before deployment. The paper's strengths are the modular pipeline supporting any Hugging Face model through vLLM, the inclusion of a new college-recommendation domain, the use of three model sizes, and the direct content-level probes (action-movie shares and Western-movie shares) that go beyond a single divergence metric. The provided 500-item lists support reproducibility. However, several load-bearing claims rest on missing baselines and absent statistical evidence, and the central bias metric is defined relative to a neutral prompt that the paper itself shows to be culturally loaded. These issues currently prevent the conclusions from being accepted at face value.

major comments (4)
  1. [§5.2.3 (H4) and §4.3] The Western-content claim lacks the catalog base rate. The movie catalog is described as highest-rated IMDb movies with English-language titles, a set almost certainly dominated by North American/European productions. Without reporting the Western share of the 500-item catalog, the neutral user's 91.3% Western-movie figure cannot be interpreted as model bias; it may simply mirror the candidate set. The same issue threatens the music catalog ('Acclaimed Music' lists). Please report the base Western proportion of each catalog and compare each persona's share to it, with a statistical test. This is load-bearing because H4 supports the abstract's 'cultural stereotypes' claim and motivates the critique that the neutral baseline is itself Western-leaning.
  2. [§5.2.3 (H1), Table 2] The claim of a 'non-linear relationship between model size and fairness' is supported by only three model sizes (1B, 4B, 12B) and no statistical testing. The means in Table 2 have standard deviations over five seeds, but there is no test of whether the 4B-vs-12B differences are reliable, no confidence intervals, and no formal comparison against monotonic alternatives. Please provide bootstrap confidence intervals or a small significance analysis (e.g., paired permutation tests across seeds) before asserting a non-monotonic relationship; with three points, the pattern may be within noise.
  3. [§5.2.3 (H3), Figure 3] The support for H3 rests entirely on a spider plot with no numeric values, no axis scales, and no table of the underlying divergence scores. The text says the orange line 'consistently lies closer from the center' and describes reductions for selected attributes, but the reader cannot verify the magnitude or variability. Please include the numeric IOU divergences with and without context for each attribute, plus seed-level variability, so the claimed mitigation is quantitatively assessable.
  4. [§4.2, Eq. (1)-(3)] The bias metric B_k(a) = 1 - Sim(I_a, I_neutral) treats the neutral prompt as the normative unbiased baseline. Section 5.2.3 (H4) then shows that this neutral baseline is heavily Western-leaning (91.3% Western movies). Consequently, a low divergence from the neutral prompt does not establish low bias; it only shows similarity to a default that the paper itself argues is culturally loaded. This circularity does not invalidate the direct genre/country measurements, but it means the headline benchmark numbers are relative to a self-defined default rather than an absolute fairness standard. Please clarify this interpretation and consider supplementing the divergence metric with catalog-based or population-based baselines.
minor comments (6)
  1. [General] Typos and formatting issues: 'compakaired' in §5.2.3, 'F an' in the Figure 3 caption, and 'V olume'/'V ersion' in the references. Please proofread.
  2. [Figure 2 and §5.2.3 (H2)] The H2 percentages (35.0%, 40.5%, 14.8%, 18.0%, 32.0%) are point estimates with no seed-level variance or significance test. Since the paper reports five seeds, please report standard deviations or error bars so the reader can assess whether the boy/girl difference is robust.
  3. [Figure 3] The spider plot is hard to read without numerical axis labels or a legend explaining which line corresponds to which condition. A table would be more informative.
  4. [§3.2] The notation P(item|a) and P(item) is informal; define the sampling interpretation precisely (e.g., whether the lists are draws from a Plackett-Luce or other distribution over the catalog).
  5. [§5.2.2] The decision to drop Llama from the main hypothesis tests is based on qualitative output instability. This is reasonable, but the paper should state explicitly that H1-H4 are claims about Gemma models only, not about Llama, to avoid over-generalization.
  6. [§6] The discussion of bias versus personalization is useful, but it should be connected to the metric definition: if some divergence is desired personalization, then B_k is not a pure fairness measure. A short paragraph reconciling this tension with the benchmark's stated goal would strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's bias measures are explicit operational definitions, and its main stereotype findings rest on direct output measurements rather than on self-referential fits or self-citations.

full rationale

The paper defines bias operationally as B_k(a) = 1 - Sim(I_k_a, I_k_Neu) and does not present this as a first-principles derivation; it is a measurement convention. The neutral prompt is used as a reference condition, and the authors explicitly investigate whether that baseline is itself culturally loaded in H4, so they do not silently assume it is unbiased. The gender-stereotype evidence (H2) compares action-movie proportions in model outputs across personas (35.0% neutral vs. 40.5% 'a boy' vs. 14.8% 'a girl'), which is a direct content measurement and does not reduce to the divergence metric. The Western-content claim (H4) is also a direct measurement of output composition; although it would be stronger with a reported catalog base rate to rule out a dataset-composition confound, the absence of that control is a validity concern, not a circular reduction—the model's selected top-20 is not logically forced by the 500-item input list. All cited prior work (FaiRLLM, CFaiRLLM, FairEval) is external to this paper's authors, and no load-bearing self-citation, imported uniqueness theorem, or ansatz-smuggling-by-citation appears. The paper's conclusions therefore have independent empirical content, and the central claims are not equivalent to the inputs by construction.

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

The ledger is mostly methodological rather than mathematical: there are no fitted constants, but the benchmark's meaning is carried by hand-chosen definitions (neutral baseline, catalog size, attribute phrasing) and by assumptions about item metadata and output parsing. The neutral baseline assumption is the heaviest: the paper's own H4 data contradicts it.

free parameters (5)
  • k (recommendation list length) = 20
    Hand-chosen in Section 5.2.1; all bias scores (IOU, SERP, PRAG) are computed on top-20 lists and therefore depend on this choice.
  • N (catalog size) = 500
    Chosen in Section 4.3 after 1,000 items exceeded the LLM context window; the item pool determines which recommendations are possible.
  • number of sampling seeds = 5
    All reported numbers are averages over 5 seeds (Section 5.2.1); no significance tests are reported, so seed count directly controls the stated precision.
  • Western countries definition = North America, Europe, Australia, New Zealand
    Defined by hand in H4 (Section 5.2.3); every Western-content percentage in Figure 4 is computed from this set.
  • gender attribute phrasing = boy, girl, male, female
    Authors show the results differ between boy/girl and male/female (Section 5.2.3, H2), so the chosen phrasing affects the stereotype conclusion.
assumptions (4)
  • ad hoc to paper The neutral prompt (no sensitive attributes) provides an unbiased baseline for measuring bias.
    Section 4.2 defines bias as 1 - Sim(I_a, I_Neu); Section 5.2.3 H4 shows the neutral user receives 91.3% Western movies, so the baseline is itself culturally loaded.
  • domain assumption Divergence from the neutral baseline indicates bias rather than legitimate personalization.
    Section 4.2 acknowledges that identical lists could mean lack of personalization; the paper nonetheless interprets divergence as bias in the headline claims.
  • domain assumption Item catalogs are representative and their metadata (genre, country of origin) is accurate.
    Section 4.3 uses API-sourced movies and music and QS-ranked colleges; genre and country-of-origin labels used in H2 and H4 are not validated.
  • domain assumption LLM outputs can be reliably parsed into the requested top-k lists.
    Section 5.2.2 documents Llama repetition artifacts that distorted outputs; the pipeline's parser is assumed to recover intended lists from noisy generations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Revealing Potential Biases in LLM-Based Recommender Systems in the Cold Start Setting." pith.science (2026). https://pith.science/paper/TIZQXQQJ

@misc{pith2026250820401,
  author       = {Pith},
  title        = {Pith review of: Revealing Potential Biases in LLM-Based Recommender Systems in the Cold Start Setting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TIZQXQQJ}},
  note         = {Machine review of arXiv:2508.20401}
}
read the original abstract

Large Language Models (LLMs) are increasingly used for recommendation tasks due to their general-purpose capabilities. While LLMs perform well in rich-context settings, their behavior in cold-start scenarios, where only limited signals such as age, gender, or language are available, raises fairness concerns because they may rely on societal biases encoded during pretraining. We introduce a benchmark specifically designed to evaluate fairness in zero-context recommendation. Our modular pipeline supports configurable recommendation domains and sensitive attributes, enabling systematic and flexible audits of any open-source LLM. Through evaluations of state-of-the-art models (Gemma 3 and Llama 3.2), we uncover consistent biases across recommendation domains (music, movies, and colleges) including gendered and cultural stereotypes. We also reveal a non-linear relationship between model size and fairness, highlighting the need for nuanced analysis.

Figures

Figures reproduced from arXiv: 2508.20401 by the authors.

Figure 1
Figure 1. Overview of the benchmark pipeline for measuring bias across datasets and sensitive attributes. The user provides datasets, sensitive attributes, and the LLM to be tested. From these inputs, the pipeline automatically generates benchmark prompts that vary by sensitive attribute (e.g., attribute A, attribute B, and neutral). The LLM is queried with these prompts and its outputs are formatted into recommendation sets … view at source ↗
Figure 2
Figure 2. Ratio of Action Movies Recommended by Gemma 3 4B Across Gender Attribute. H3: Adding context to a user mitigates bias This hypothesis proposes that incorporating relevant user preferences into the prompt, alongside sensitive attributes, can diminish the influence of those sensitive attributes, leading to reduced bias. We conducted this test using Gemma 3 12B on the Movie dataset, given its previously observed high s… view at source ↗
Figure 3
Figure 3. Impact of ’Action Movie Fan’ Context on IOU Divergence in Gemma 3 12B (Movie Dataset) A general trend emerges: IOU divergence scores are lower across many sensitive attributes when the ’action movie fan’ context is included. In [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Ratio of Western Movies Recommended by Gemma 3 12B Across Different Attribute. 6 Discussion Limitations and Extensions An alternative analysis could reverse the current setup, ranking users for a given item instead of ranking items for a user. This perspective could re…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Invisible to the Machine: Auditing AI Restaurant, Cafe, and Bar Recommendation Against a Complete Market Census

    cs.IR 2026-08 conditional novelty 7.0 of 10

    Across 2,208 AI queries, 85.6% of 4,776 food and drink venues were never recommended; entry into answers tracks documentation, while star rating only affects rank among recommended venues.

  2. Critical Acclaim Orientation in Large Language Models: Evidence from Film Preference Elicitation

    cs.AI 2026-08 conditional novelty 7.0 of 10

    LLMs from four families systematically favor critically acclaimed but commercially obscure films over commercial blockbusters, and the preference strengthens with model scale.

Reference graph

Works this paper leans on

32 extracted references · 28 canonical work pages · cited by 2 Pith papers

  1. [1]

    Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M

    Jason Wei, Maarten Bosma, Vincent Y . Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, and Quoc V . Le. Finetuned language models are zero-shot learners, 2022

    work page 2022

  2. [2]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Nee- lakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Lit...

    2020

  3. [3]

    Text2playlist: Generating personalized playlists from text on deezer, 2025

    Mathieu Delcluze, Antoine Khoury, Clémence Vast, Valerio Arnaudo, Léa Briand, Walid Bendada, and Thomas Bouabça. Text2playlist: Generating personalized playlists from text on deezer, 2025

    work page 2025

  4. [4]

    Recommendation as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5), 2023

    Shijie Geng, Shuchang Liu, Zuohui Fu, Yingqiang Ge, and Yongfeng Zhang. Recommendation as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5), 2023

    work page 2023

  5. [5]

    Is chatgpt fair for recommenda- tion? evaluating fairness in large language model recommendation

    Jizhi Zhang, Keqin Bao, Yang Zhang, Wenjie Wang, Fuli Feng, and Xiangnan He. Is chatgpt fair for recommenda- tion? evaluating fairness in large language model recommendation. InProceedings of the 17th ACM Conference on Recommender Systems, RecSys ’23, page 993–999. ACM, September 2023

    work page 2023

  6. [6]

    Faireval: A benchmark for evaluating user-level fairness in large language model-based recommender systems

    Ziwei Wan, Jiaxin Wu, Weiqing Liu, Leyan Zhou, Chen Zhu, Bin Hu, Zhiqiang Liu, Yang Liu, and Jing Liu. Faireval: A benchmark for evaluating user-level fairness in large language model-based recommender systems. In Proceedings of the ACM Web Conference 2024, pages 1323–1334, 2024

    work page 2024

  7. [7]

    Cfairllm: Controlling consumer fairness in large language model-based recommender systems

    Yihan Wang, Yupeng Zhang, Xiangnan He, and Tat-Seng Chua. Cfairllm: Controlling consumer fairness in large language model-based recommender systems. InProceedings of the 17th ACM International Conference on Web Search and Data Mining, pages 807–817, 2024

    work page 2024

  8. [8]

    Large language models are competitive near cold-start recommenders for language- and item-based preferences, 07 2023

    Scott Sanner, Krisztian Balog, Filip Radlinski, Ben Wedin, and Lucas Dixon. Large language models are competitive near cold-start recommenders for language- and item-based preferences, 07 2023

    work page 2023

Show all 32 references
  1. [9]

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

    Keqin Bao, Jizhi Zhang, Yang Zhang, Wenjie Wang, Fuli Feng, and Xiangnan He. Tallrec: An effective and efficient tuning framework to align large language model with recommendation. InProceedings of the 17th ACM Conference on Recommender Systems, RecSys ’23, page 1007–1014. ACM...

    2023

  2. [10]

    Filterllm: Text-to-distribution llm for billion-scale cold-start recommendation, 2025

    Ruochen Liu, Hao Chen, Yuanchen Bei, Zheyu Zhou, Lijia Chen, Qijie Shen, Feiran Huang, Fakhri Karray, and Senzhang Wang. Filterllm: Text-to-distribution llm for billion-scale cold-start recommendation, 2025

    2025

  3. [11]

    Stereoset: Measuring stereotypical bias in pretrained language models

    Moin Nadeem, Anna Bethke, and Siva Reddy. Stereoset: Measuring stereotypical bias in pretrained language models. InProceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing...

    2021

  4. [12]

    Alicia Parrish, Angelica Chen, Nikita Nangia, Vishakh Padmakumar, Jason Phang, Jana Thompson, Phu Mon Htut, and Samuel R. Bowman. Bbq: A hand-built bias benchmark for question answering, 2022

    2022

  5. [13]

    Realtoxicityprompts: Evaluating neural toxic degeneration in language models

    Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A Smith. Realtoxicityprompts: Evaluating neural toxic degeneration in language models. InFindings of the Association for Computational Linguistics: EMNLP 2020, pages 3356–3369. Association for Computational Li...

    2020

  6. [14]

    ToxiGen: A large-scale machine-generated dataset for adversarial and implicit hate speech detection

    Thomas Hartvigsen, Saadia Gabriel, Hamid Palangi, Maarten Sap, Dipankar Ray, and Ece Kamar. ToxiGen: A large-scale machine-generated dataset for adversarial and implicit hate speech detection. In Smaranda Muresan, Preslav Nakov, and Aline Villavicencio, editors,Proceedings of ...

    2022

  7. [15]

    Ethics and governance of artificial intelligence: Evidence from a survey of machine learning researchers

    Dan Hendrycks, Mantas Mazeika, Andy Zou, Maya Musser, Jacob Zhu, Nelson F Li, Dawn Song, and Jacob Steinhardt. Ethics and governance of artificial intelligence: Evidence from a survey of machine learning researchers. InProceedings of the AAAI/ACM Conference on AI, Ethics, and ...

    2021

  8. [16]

    Eagle: Ethical dataset given from real interactions, 2024

    Masahiro Kaneko, Danushka Bollegala, and Timothy Baldwin. Eagle: Ethical dataset given from real interactions, 2024

    2024

  9. [17]

    Towards understanding and mitigating social biases in language models

    Percy Liang, Chunyuan Li, Charles Zheng, et al. Towards understanding and mitigating social biases in language models. InInternational Conference on Machine Learning, pages 6565–6576. PMLR, 2021

    2021

  10. [18]

    Certifying counterfactual bias in llms, 2025

    Isha Chaudhary, Qian Hu, Manoj Kumar, Morteza Ziyadi, Rahul Gupta, and Gagandeep Singh. Certifying counterfactual bias in llms, 2025. 10 Revealing Potential Biases in LLM-Based Recommender Systems in the Cold Start Setting

    2025

  11. [19]

    Controlling popularity bias in recommender systems

    Himan Abdollahpouri, Robin Burke, and Bamshad Mobasher. Controlling popularity bias in recommender systems. InProceedings of the Eleventh ACM Conference on Recommender Systems, pages 42–46, 2017

    2017

  12. [20]

    Evaluating the impact of interaction sparsity on machine learning for exposure bias mitigation in recommender systems

    Fernando Diaz, Bhaskar Mitra, Michael D Ekstrand, and Asia J Biega. Evaluating the impact of interaction sparsity on machine learning for exposure bias mitigation in recommender systems. InProceedings of the 43rd International ACM SIGIR Conference on Research and Development i...

    2020

  13. [21]

    Multisided fairness for recommendation.arXiv preprint arXiv:1707.00093, 2017

    Robin Burke. Multisided fairness for recommendation.arXiv preprint arXiv:1707.00093, 2017

    2017 arXiv

  14. [22]

    Beyond accuracy: evaluating recommender systems by coverage and serendipity

    Mouzhi Ge, Clélia Delgado-Battenfeld, and Dietmar Jannach. Beyond accuracy: evaluating recommender systems by coverage and serendipity. InProceedings of the fourth ACM conference on Recommender systems, pages 257–260, 2010

    2010

  15. [23]

    Calibrated recommendations

    Harald Steck. Calibrated recommendations. InProceedings of the 12th ACM conference on recommender systems, pages 154–162, 2018

    2018

  16. [24]

    Beyond parity: Fairness objectives for collaborative filtering

    Sirui Yao and Bert Huang. Beyond parity: Fairness objectives for collaborative filtering. InAdvances in neural information processing systems, volume 30, 2017

    2017

  17. [25]

    A survey on fairness-aware recommender systems, 2023

    Di Jin, Luzhi Wang, He Zhang, Yizhen Zheng, Weiping Ding, Feng Xia, and Shirui Pan. A survey on fairness-aware recommender systems, 2023

    2023

  18. [26]

    Gonzalez, Hao Zhang, and Ion Stoica

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. InProceedings of the ACM SIGOPS 29th Symposium on Operating Systems ...

    2023

  19. [27]

    Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models, 2020

    2020

  20. [28]

    Unmasking gender bias in recommendation systems and enhancing category-aware fairness

    Tahsin Alamgir Kheya, Mohamed Reda Bouadjenek, and Sunil Aryal. Unmasking gender bias in recommendation systems and enhancing category-aware fairness. InProceedings of the ACM on Web Conference 2025, WWW ’25, page 5127–5138, New York, NY , USA, 2025. Association for Computing ...

    2025

  21. [29]

    Challenging fairness: A comprehensive exploration of bias in llm-based recommendations

    Shahnewaz Karim Sakib and Anindya Bijoy Das. Challenging fairness: A comprehensive exploration of bias in llm-based recommendations. In2024 IEEE International Conference on Big Data (BigData), pages 1585–1592. IEEE, 2024

    2024

  22. [30]

    Lian Xiaoli, and Muhammad Mirajul Islam

    Chandan Kumar Sah, Dr. Lian Xiaoli, and Muhammad Mirajul Islam. Unveiling bias in fairness evaluations of large language models: A critical literature review of music and movie recommendation systems.Unveiling Bias in Fairness Evaluations of Large Language Models: A Critical L...

    2024

  23. [31]

    DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei F...

    2025

  24. [32]

    Unveiling and mitigating bias in large language model recommendations: A path to fairness, 2024

    Anindya Bijoy Das and Shahnewaz Karim Sakib. Unveiling and mitigating bias in large language model recommendations: A path to fairness, 2024. 12

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.