Pith. sign in

REVIEW 4 major objections 5 minor 47 references

Web-Browsing LLMs Can Access Social Media Profiles and Infer User Demographics

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Web-browsing LLMs can retrieve X profiles and infer user demographics from usernames alone, with accuracy that varies by model and attribute.

desk verdict Plausible new capability, but the headline accuracy rests on a memorization confound; the synthetic data are too thinned out to carry the load. read the letter →

arxiv 2507.12372 v1 pith:OJ2HJI4N submitted 2025-07-16 cs.CL

Meysam Alizadeh , Fabrizio Gilardi , Zeynab Samei , Mohsen Mosleh This is my paper
classification cs.CL
keywords largelanguagemodelswebbrowsingsocialmediaprofilingdemographicinferenceX(Twitter)accountsprivacyriskscomputationalscienceusermodeling
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 asks whether web-browsing large language models can do what older classifiers could not: start from a bare X (formerly Twitter) handle, retrieve that person's live profile content, and infer their age, gender, socioeconomic status, and political orientation. The authors report that GPT-4o and GPT-o3 do this with accuracy well above chance on a 1,384-user survey dataset, and that even a smaller open model attempts the retrieval. A controlled experiment with 27 synthetic accounts shows the models draw on profile pictures, bios, and follow lists, and that they fall back on names even for suspended accounts, which can introduce gender and political bias. The paper positions this as a new zero-data profiling capability that could sustain computational social science after API restrictions or enable mass surveillance and targeting, and it recommends restricting the capability in public products.

What carries the argument

The operative mechanism is the web-browsing retrieval loop: the model receives a prompt containing only the account handle or profile link, internally fetches the live X page, and conditions its demographic answer on the fetched content including posts, bio, profile picture, replies, retweets, and follow lists. The paper's best-performing configuration is the 'Chatbot-Handle' prompt, which asks the model to complete the sentence "Based on tweets posted by handle, given the options [...], I think the [attribute] of this user is:" and which matched or outperformed other prompt and identifier combinations across models and datasets. The controlled synthetic design is what lets the authors attribute inferences to specific profile components: one-third of accounts had picture plus bio, one-third only bio, one-third only picture, and half followed 25 politically aligned accounts, so accuracy differences across those conditions reveal which cues the models actually use.

What would settle it

Take a fresh cohort of X accounts created after every relevant model's training cutoff, with known demographics and fully public content; prompt the same models by username before and after making the accounts crawlable. If accuracy collapses when the accounts cannot be retrieved, the capability is genuine browsing; if accuracy persists on blocked or nonexistent handles, the dominant mechanism is memorized or hallucinated content.

Watch

Extended reading notes

Core claim

The central claim is that browsing-enabled LLMs constitute a working profiler: given only a username, they access the account's posts, bio, replies, retweets, and profile picture, and predict demographic attributes with moderate-to-high accuracy. On the 1,384-user survey dataset, GPT-4o reaches 72–88 percent accuracy on age, gender, and socioeconomic status, and GPT-o3 reaches 90 percent on age; political orientation is lower but still mostly above chance. On the 27 usable synthetic accounts, accuracy is lower and is treated by the authors as a lower bound, with gender the most reliably inferred attribute and age the hardest. The paper also claims that the mechanism is visible in the synthetic experiments: models rely more on names and profile pictures than on textual history, which produces bias against minimal-activity accounts and, for GPT-4o, against conservative users.

Load-bearing premise

The survey-dataset results count as evidence for live web browsing only if the models are not instead recalling profiles they have already memorized from training data; the paper itself notes the 2018 accounts were likely used to train Grok and possibly ChatGPT, and no control condition separates browsing from memory.

Editorial extensions

If this is right

  • Demographic inference from a username is now a live, low-cost operation, so the privacy risk is no longer hypothetical: anyone with API access can profile users at scale without an API feed.
  • Researchers in the post-API era can use browsing LLMs to reconstruct user-level attributes for studies that would otherwise lack data, provided ethical review and consent are handled.
  • Because models infer gender and ideology from names and pictures even when the account is suspended, profiles with minimal activity inherit systematic misclassification bias.
  • The accuracy gap between synthetic and survey accounts implies that real, organically maintained accounts give the models richer signals, so performance in natural deployments should be expected to be higher, not lower.

Reading between the lines

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

  • The paper's memorization caveat suggests a strong test the authors did not run: if accounts are suspended or blocked between prompt and inference and accuracy persists, the survey results would indicate training-data recall rather than live browsing.
  • The same mechanism should transfer to other platforms with public profiles, which would widen both the research upside and the surveillance surface; a systematic multi-platform benchmark is the natural next step.
  • The reliance on names and profile pictures for gender and ideology implies that pseudonymous handles or neutral avatars could degrade inference accuracy, a testable privacy-preserving design.
  • If political-orientation accuracy stays low on real accounts, targeted political advertising from usernames may be less effective than feared, whereas gender and age targeting would remain the sharper risk.
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. This paper investigates whether web-browsing large language models (GPT-4o, GPT-o3, and Llama-3-8B-Web) can, given only an X/Twitter handle, retrieve profile content and infer a user's age, gender, socioeconomic status, and political orientation. The authors use two datasets: 48 synthetic accounts created for the study (27 remain after suspensions) and 1,384 accounts from a 2018 survey with self-reported demographics. They report above-chance accuracy on the survey dataset, explore which profile signals the models rely on using the synthetic accounts, and discuss privacy, security, and policy implications. The central claim is that web-browsing LLMs enable zero-data social media profiling with reasonable accuracy.

Significance. If the capability claim is established, this is a novel and important result for computational social science and for privacy research: it would show that an LLM with live web access can profile individuals from a username alone, with no API access. The paper has clear strengths: a controlled synthetic-account design with pre-specified demographic profiles, careful attention to research ethics (private accounts, no interaction with users, anonymized survey data), honest reporting of limitations such as account suspension and lower-bound interpretation, and a useful mechanism analysis. The synthetic-account portion is a genuinely interesting probe of how models use profile pictures, bios, and following lists. However, the quantitative headline results rest primarily on the survey dataset, which is acknowledged to overlap with the models' training data, and no control separates live browsing from parametric memory. The current evidence is therefore insufficient to support the paper's central claim at the stated strength.

major comments (4)
  1. [§3.1, §4.2] The survey-dataset result is confounded by training-data memorization, and this is load-bearing for the paper's central claim. Section 3.1 explicitly states that the 2018 accounts 'have already been used to train Grok as well as, most likely, other models such as ChatGPT.' Because the prompts in Table 1 ask the model to answer 'Based on tweets posted by handle/link,' the model can answer from parametric memory of those tweets without performing any live retrieval. The paper reports no control condition with web browsing disabled, no retrieval logs, and no independent verification that the models actually fetched current profile content. Consequently, the high accuracies in Figure 2 and Tables 7–8 do not establish that web browsing enables the profiling capability; they may reflect memorized training data. This is the central unsupported claim of the paper.
  2. [§4.1, §3.1] The synthetic dataset, which avoids the training-data confound, is too small and too attrition-prone to support the mechanism claims. Of the 48 accounts, 21 were suspended, leaving 27, and all 16 accounts in the no-bio condition were suspended. This means the controlled manipulation of bios cannot be evaluated at all, and the comparisons in Figure 1b–1d (profile picture, following, suspended vs. non-suspended) rest on subgroups of the 27 accounts with no statistical testing. The paper repeatedly notes the small sample and says results are exploratory, but Section 4.1 presents these comparisons as findings (e.g., 'the presence of a profile picture improves gender prediction accuracy across all three models'). As presented, the evidence cannot distinguish true signals from noise, and the loss of the no-bio condition removes a key part of the intended design.
  3. [§4.1, §4.2, Appendix C] The choice of the Chatbot-Handle prompting combination is post hoc and likely inflates reported accuracy. The paper states that no single prompt-identifier combination consistently outperformed others, then selects Chatbot-Handle for the main results. However, Appendix C shows that other combinations often perform better: for GPT-4o on the survey dataset, System-Handle gives age accuracy 0.83 vs. 0.72 for Chatbot-Handle, and User-Handle gives socioeconomic accuracy 0.90 vs. 0.88. For GPT-o3 on the synthetic dataset, System-Link gives political orientation 0.62 vs. 0.52 for Chatbot-Handle. Reporting the best-performing combination without pre-registration or multiple-comparison correction makes the headline accuracy numbers optimistic. The authors should either justify the selection criterion on a priori grounds or report all six combinations as primary analyses.
  4. [§4.2, §3.3] Accuracy comparisons to 'random chance' and significance claims are not statistically substantiated. After category collapsing, chance levels are 50% for gender, 25% for age (4 classes), and 33% for socioeconomic status and political orientation (3 classes). GPT-o3's political accuracy of 0.42 is only 9 points above chance, yet Figure 2a describes accuracies as 'substantially above random chance' without confidence intervals or hypothesis tests. The text also uses 'significantly' for model comparisons and for the conservative-user decline (Figure 2c, 2d), but no significance tests, confidence intervals, or effect sizes are reported for any accuracy value. This is a load-bearing issue because the paper's comparative claims (e.g., which model is best, which groups are biased) are presented as findings without inferential support.
minor comments (5)
  1. [Figure 1 caption] The figure caption is inconsistent with the panel labels: the caption lists (c) as 'accounts with and without bios' while the panel itself is labeled 'Following Vs. Non-Following,' and panels (e) and (f) are not described in the caption. Please align the caption with the actual panel content.
  2. [Throughout] There are several typos and inconsistencies: 'perfromance' in Section 2; 'and and' in Section 4.2; 'the different is not significant' should be 'the difference is not significant'; 'GPT-3.5 (o3)' should be 'GPT-o3'; 'Amozon' in the Appendix tweets; and 'underrate' in a tweet appears truncated.
  3. [Section 3.1] The paragraph on IRB approval argues that the survey data are publicly available and that X's terms permit use for model training, but the models are not 'trained' by the research; the paper uses them in inference. This does not affect the technical results, but the framing could be tightened.
  4. [Section 4.1, 'Suspended Accounts'] The claim that suspended accounts have 'only the profile picture and screen name visible' is used to infer that models rely on names and pictures, but this inference depends on the assumption that the models do not have cached or parametrically memorized content from before suspension. Given the training-data concern raised elsewhere, this assumption should be acknowledged.
  5. [Appendix A] The system prompt in Figure 3 contains a grammatical error ('You live is the USA') and defines age ranges that are inconsistent with the categories in Table 2 (e.g., 'below 14 years old' vs. 'Child (between 13 and 18)'). Please harmonize these definitions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the memorization confound in the survey dataset is a validity issue, not a by-construction reduction.

full rationale

Assigning 0. The paper's claimed derivation chain (web-browsing LLM given a handle, retrieves profile content, then infers demographics) does not reduce to its own inputs through any equation or fitted parameter. No parameter is fitted to the survey labels, no result is defined in terms of another result, and no load-bearing argument depends on a self-citation. The synthetic dataset uses known labels by design, but that is a benchmark construction, not a circular derivation; the fact that GPT-4o both generated the persona content and later inferred the same demographics is a possible same-model bias, not an equivalence between prediction and input. The survey dataset's main threat is memorization: Section 3.1 states that all accounts 'have already been used to train Grok as well as, most likely, other models such as ChatGPT,' and the study lacks a no-browsing control to separate live retrieval from parametric recall. This is a real confound for the web-access capability claim and should be weighed as correctness/validity risk, but it is not one of the enumerated circularity patterns: recalling training data is not the same as predicting from a fitted parameter renamed as a prediction, and the demographic labels are not encoded in the username input by construction. Self-citations in the paper ([1], [4], [15], [25]) are contextual—data reuse, prior annotation results, and risk examples—and do not force the conclusions. The honest finding is therefore no significant circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the assumption that observed accuracy is caused by live browsing and on the representativeness of the small synthetic sample. The selection of the Chatbot-Handle prompt and the collapsing of categories are post-hoc decisions that can inflate accuracy. No new entities are introduced.

free parameters (3)
  • Prompt-identifier combination selection = Chatbot-Handle
    The authors evaluated six prompt and identifier combinations and report only the one that showed relatively robust performance, a post-hoc selection that can inflate reported accuracy.
  • Category collapsing for survey data = Three classes for SES and political orientation
    The five-category survey answers were collapsed to three classes to match the synthetic dataset, changing the chance baseline and potentially inflating accuracy if the middle class is dominant.
  • Account maturity threshold = 15 days and 30 tweets
    The synthetic accounts were active for 15 days with 30 tweets, and the authors treat this as a lower bound, but the threshold is arbitrary and affects how signals are available to the browsing LLM.
assumptions (4)
  • domain assumption LLM web-browsing correctly retrieves the live X profile content
    The paper provides example screenshots but does not systematically verify that each classification was based on live retrieval rather than training data or hallucination.
  • domain assumption The synthetic accounts' tweets accurately reflect the assigned demographic profiles
    GPT-4o generated the tweets from prompts containing the true demographics, so content is stereotypical by construction, which may make inference easier and may drive the bias findings.
  • domain assumption The 2018 survey users' handles are stable and still belong to the same people
    The paper assumes that the 1,384 active handles correspond to the same users who participated in the 2018 survey, though handles can be abandoned, sold, or recycled.
  • domain assumption Self-reported demographics in the survey are accurate
    The ground truth for the survey dataset relies on self-report from 2018, which may contain errors or changes over time.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Web-Browsing LLMs Can Access Social Media Profiles and Infer User Demographics." pith.science (2026). https://pith.science/paper/OJ2HJI4N

@misc{pith2026250712372,
  author       = {Pith},
  title        = {Pith review of: Web-Browsing LLMs Can Access Social Media Profiles and Infer User Demographics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OJ2HJI4N}},
  note         = {Machine review of arXiv:2507.12372}
}
read the original abstract

Large language models (LLMs) have traditionally relied on static training data, limiting their knowledge to fixed snapshots. Recent advancements, however, have equipped LLMs with web browsing capabilities, enabling real time information retrieval and multi step reasoning over live web content. While prior studies have demonstrated LLMs ability to access and analyze websites, their capacity to directly retrieve and analyze social media data remains unexplored. Here, we evaluate whether web browsing LLMs can infer demographic attributes of social media users given only their usernames. Using a synthetic dataset of 48 X (Twitter) accounts and a survey dataset of 1,384 international participants, we show that these models can access social media content and predict user demographics with reasonable accuracy. Analysis of the synthetic dataset further reveals how LLMs parse and interpret social media profiles, which may introduce gender and political biases against accounts with minimal activity. While this capability holds promise for computational social science in the post API era, it also raises risks of misuse particularly in information operations and targeted advertising underscoring the need for safeguards. We recommend that LLM providers restrict this capability in public facing applications, while preserving controlled access for verified research purposes.

Figures

Figures reproduced from arXiv: 2507.12372 by the authors.

Figure 1
Figure 1. Classification accuracy on the synthetic account dataset. (a) Comparison of classification performance by GPT-4o, GPT-o3, and LLaMA-3-8B-Web across four attributes: age, gender, political orientation, and socioeconomic status. (b) Accuracy comparison between accounts with and without profile pictures. (c) Accuracy comparison between accounts with and without bios. (d) Accuracy comparison between female and male acco… view at source ↗
Figure 2
Figure 2. Classification accuracy on the survey dataset. (a) Comparison of classification performance by GPT-4o and GPT-o3 across four attributes: age, gender, political orientation, and socioeconomic status. (b) Accuracy comparison between female and male accounts. predicting the demographics of male and female users. GPT-4o does not exhibit notable gender bias in inferring age, gender, or socioeconomic status from users’ ac… view at source ↗
Figure 3
Figure 3. The default system prompt used to generate social media posts across various personas. System prompt = """ Your gender is {gender}. Your age is {age}. Your political orientation is {political orientation}. Your educational level is {education}. Your socioeconomic status is {socioeco}. Generate 15 social media posts to be posted on your X (Twitter) account. Make sure the tweets reflect your gender, age, political ori… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Prompt for generating social media posts for a specific persona. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Sample of the 48 dummy X (Twitter) accounts created for experimental setup. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: A sample response from Mistral when asked to predict the political orientation of the first [PITH_FULL_IMAGE:figures/full_fig_p034_6.png]
Figure 7
Figure 7. Figure 7: Sample of LLMs responses to queries about accessing replies, retweets, and historical data. 35 [PITH_FULL_IMAGE:figures/full_fig_p035_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 30 canonical work pages

  1. [1]

    Data marketplaces can increase the willingness to share social media data at low prices

    Meysam Alizadeh and Fabrizio Gilardi. Data marketplaces can increase the willingness to share social media data at low prices. arXiv preprint arXiv:2506.16618, 2025

    work page Pith review arXiv 2025

  2. [2]

    Content-based features predict social media influence operations

    Meysam Alizadeh, Jacob N Shapiro, Cody Buntain, and Joshua A Tucker. Content-based features predict social media influence operations. Science advances, 6(30):eabb5824, 2020

    work page 2020

  3. [3]

    Tokenization of social media engagements increases the sharing of false (and other) news but penalization moderates it

    Meysam Alizadeh, Emma Hoes, and Fabrizio Gilardi. Tokenization of social media engagements increases the sharing of false (and other) news but penalization moderates it. Scientific Reports, 13(1):13703, 2023

    work page 2023

  4. [4]

    Simple prompt injection attacks can leak personal data observed by llm agents during task execution

    Meysam Alizadeh, Zeynab Samei, Daria Stetsenko, and Fabrizio Gilardi. Simple prompt injection attacks can leak personal data observed by llm agents during task execution. arXiv preprint arXiv:2506.01055, 2025

    arXiv 2025

  5. [5]

    Principles and Guidance for Human Subjects Research,

    American Political Science Association. Principles and Guidance for Human Subjects Research,

    work page

  6. [6]

    Can generative ai improve social science? Proceedings of the National Academy of Sciences, 121(21):e2314021121, 2024

    Christopher A Bail. Can generative ai improve social science? Proceedings of the National Academy of Sciences, 121(21):e2314021121, 2024

    work page 2024

  7. [7]

    Do we need a social media accelerator? 2023

    Christopher A Bail, D Sunshine Hillygus, Alexander V olfovsky, Max Allamong, Fatima Alqa- bandi, Diana ME Jordan, Graham Tierney, Christina Tucker, Andrew Trexler, and Austin van Loon. Do we need a social media accelerator? 2023

    work page 2023

  8. [8]

    Latent dirichlet allocation

    David M Blei, Andrew Y Ng, and Michael I Jordan. Latent dirichlet allocation. Journal of machine Learning research, 3(Jan):993–1022, 2003

    work page 2003

Show all 47 references
  1. [9]

    Designing a dashboard for transparency and control of conversational ai

    Yida Chen, Aoyu Wu, Trevor DePodesta, Catherine Yeh, Kenneth Li, Nicholas Castillo Marin, Oam Patel, Jan Riecke, Shivam Raval, Olivia Seow, et al. Designing a dashboard for transparency and control of conversational ai. arXiv preprint arXiv:2406.07882, 2024

    2024 arXiv

  2. [10]

    Detecting hate speech with gpt-3

    Ke-Li Chiu, Annie Collins, and Rohan Alexander. Detecting hate speech with gpt-3. arXiv preprint arXiv:2103.12407, 2021

    2021 arXiv

  3. [11]

    Leveraging large language models for fact-checking farsi news headlines

    Shirin Dehghani, Mohammadmasiha Zahedivafa, Zahra Baghshahi, Darya Zare, Sara Yari, Zeynab Samei, Mohammadhadi Aliahmadi, Mahdis Abbasi, Sara Mirzamojtahedi, Sarvenaz Ebrahimi, et al. Leveraging large language models for fact-checking farsi news headlines. In Multidisciplinary...

    2024

  4. [12]

    Enhancing job recommendation through llm-based generative adversarial networks

    Yingpeng Du, Di Luo, Rui Yan, Xiaopei Wang, Hongzhi Liu, Hengshu Zhu, Yang Song, and Jie Zhang. Enhancing job recommendation through llm-based generative adversarial networks. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 8363–8371, 2024

    2024

  5. [13]

    Social bot detection in the age of chatgpt: Challenges and opportunities

    Emilio Ferrara. Social bot detection in the age of chatgpt: Challenges and opportunities. First Monday, 2023

    2023

  6. [14]

    Computational research in the post-api age

    Deen Freelon. Computational research in the post-api age. Political Communication, 35(4): 665–668, 2018

    2018

  7. [15]

    Chatgpt outperforms crowd workers for text-annotation tasks

    Fabrizio Gilardi, Meysam Alizadeh, and Maël Kubli. Chatgpt outperforms crowd workers for text-annotation tasks. Proceedings of the National Academy of Sciences, 120(30):e2305016120, 2023

    2023

  8. [16]

    Distributional structure

    Zellig S Harris. Distributional structure. Word, 10(2-3):146–162, 1954

    1954

  9. [17]

    A user profile modeling method based on word2vec

    Jianqiao Hu, Feng Jin, Guigang Zhang, Jian Wang, and Yi Yang. A user profile modeling method based on word2vec. In 2017 IEEE International Conference on Software Quality, Reliability and Security Companion (QRS-C), pages 410–414. IEEE, 2017. 11

    2017

  10. [18]

    Human heuristics for ai-generated language are flawed

    Maurice Jakesch, Jeffrey T Hancock, and Mor Naaman. Human heuristics for ai-generated language are flawed. Proceedings of the National Academy of Sciences, 120(11):e2208839120, 2023

    2023

  11. [19]

    Sentimentgpt: Exploiting gpt for advanced sentiment analysis and its departure from current machine learning

    Kiana Kheiri and Hamid Karimi. Sentimentgpt: Exploiting gpt for advanced sentiment analysis and its departure from current machine learning. arXiv preprint arXiv:2307.10234, 2023

    2023 arXiv

  12. [20]

    Ai safety on whose terms?, 2023

    Seth Lazar and Alondra Nelson. Ai safety on whose terms?, 2023

    2023

  13. [21]

    A survey on representation learning for user modeling

    Sheng Li and Handong Zhao. A survey on representation learning for user modeling. In Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence, pages 4997–5003, 2021

    2021

  14. [22]

    A first look at llm-powered generative news recommendation

    Qijiong Liu, Nuo Chen, Tetsuya Sakai, and Xiao-Ming Wu. A first look at llm-powered generative news recommendation. CoRR, 2023

    2023

  15. [23]

    Once: Boosting content-based recommendation with both open-and closed-source large language models

    Qijiong Liu, Nuo Chen, Tetsuya Sakai, and Xiao-Ming Wu. Once: Boosting content-based recommendation with both open-and closed-source large language models. In Proceedings of the 17th ACM International Conference on Web Search and Data Mining, pages 452–461, 2024

    2024

  16. [24]

    Gpt-4v (ision) as a social media analysis engine

    Hanjia Lyu, Jinfa Huang, Daoan Zhang, Yongsheng Yu, Xinyi Mou, Jinsheng Pan, Zhengyuan Yang, Zhongyu Wei, and Jiebo Luo. Gpt-4v (ision) as a social media analysis engine. ACM Transactions on Intelligent Systems and Technology, 16(3):1–54, 2025

    2025

  17. [25]

    Cognitive reflection correlates with behavior on twitter

    Mohsen Mosleh, Gordon Pennycook, Antonio A Arechar, and David G Rand. Cognitive reflection correlates with behavior on twitter. Nature communications, 12(1):921, 2021

    2021

  18. [26]

    Ai snake oil: What artificial intelligence can do, what it can’t, and how to tell the difference

    Arvind Narayanan and Sayash Kapoor. Ai snake oil: What artificial intelligence can do, what it can’t, and how to tell the difference. In AI Snake Oil. Princeton University Press, 2024

    2024

  19. [27]

    Adversarial search engine optimiza- tion for large language models

    Fredrik Nestaas, Edoardo Debenedetti, and Florian Tramèr. Adversarial search engine optimiza- tion for large language models. arXiv preprint arXiv:2406.18382, 2024

    2024 arXiv

  20. [28]

    Fine-tuning llama 2 large lan- guage models for detecting online sexual predatory chats and abusive texts

    Thanh Thi Nguyen, Campbell Wilson, and Janis Dalins. Fine-tuning llama 2 large lan- guage models for detecting online sexual predatory chats and abusive texts. arXiv preprint arXiv:2308.14683, 2023

    2023 arXiv

  21. [29]

    Why open-source generative ai models are an ethical way forward for science

    Arthur Spirling. Why open-source generative ai models are an ethical way forward for science. Nature, 616(7957):413–413, 2023

    2023

  22. [30]

    Bert4rec: Sequential recommendation with bidirectional encoder representations from transformer

    Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang. Bert4rec: Sequential recommendation with bidirectional encoder representations from transformer. InPro- ceedings of the 28th ACM international conference on information and knowledge management, pages...

    2019

  23. [31]

    User modeling in the era of large language models: Current research and future directions

    Zhaoxuan Tan and Meng Jiang. User modeling in the era of large language models: Current research and future directions. arXiv preprint arXiv:2312.11518, 2023

    2023 arXiv

  24. [32]

    The only way is ethics: A guide to ethical re- search with large language models

    Eddie L Ungless, Nikolas Vitsakis, Zeerak Talat, James Garforth, Björn Ross, Arno Onken, Atoosa Kasirzadeh, and Alexandra Birch. The only way is ethics: A guide to ethical re- search with large language models. In Proceedings of the 31st International Conference on Computation...

    2025

  25. [33]

    Online human-bot interactions: Detection, estimation, and characterization

    Onur Varol, Emilio Ferrara, Clayton Davis, Filippo Menczer, and Alessandro Flammini. Online human-bot interactions: Detection, estimation, and characterization. In Proceedings of the international AAAI conference on web and social media, volume 11, pages 280–289, 2017

    2017

  26. [34]

    Clickbait detection via large language models

    Han Wang, Yi Zhu, Ye Wang, Yun Li, Yunhao Yuan, and Jipeng Qiang. Clickbait detection via large language models. arXiv preprint arXiv:2306.09597, 2023

    2023 arXiv

  27. [35]

    Zero-shot next-item recommendation using large pretrained language models

    Lei Wang and Ee-Peng Lim. Zero-shot next-item recommendation using large pretrained language models. arXiv preprint arXiv:2304.03153, 2023. 12

    2023 arXiv

  28. [36]

    Demographic inference and representative population estimates from multilingual social media data

    Zijian Wang, Scott Hale, David Ifeoluwa Adelani, Przemyslaw Grabowicz, Timo Hartman, Fabian Flöck, and David Jurgens. Demographic inference and representative population estimates from multilingual social media data. In The World Wide Web Conference, pages 2056–2067. ACM, 2019

    2019

  29. [37]

    Chain of thought prompting elicits reasoning in large language models

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. Chain of thought prompting elicits reasoning in large language models. arXiv preprint arXiv:2201.11903, 2022

    2022 arXiv

  30. [38]

    Taxonomy of risks posed by language models

    Laura Weidinger, Jonathan Uesato, Maribeth Rauh, Conor Griffin, Po-Sen Huang, John Mellor, Amelia Glaese, Myra Cheng, Borja Balle, Atoosa Kasirzadeh, et al. Taxonomy of risks posed by language models. In Proceedings of the 2022 ACM conference on fairness, accountability, and t...

    2022

  31. [39]

    Towards open-world recommendation with knowledge augmentation from large language models

    Yunjia Xi, Weiwen Liu, Jianghao Lin, Xiaoling Cai, Hong Zhu, Jieming Zhu, Bo Chen, Ruiming Tang, Weinan Zhang, and Yong Yu. Towards open-world recommendation with knowledge augmentation from large language models. In Proceedings of the 18th ACM Conference on Recommender System...

    2024

  32. [40]

    Palr: Personalization aware llms for recommendation

    Fan Yang, Zheng Chen, Ziyan Jiang, Eunah Cho, Xiaojiang Huang, and Yanbin Lu. Palr: Personalization aware llms for recommendation. arXiv preprint arXiv:2305.07622, 2023

    2023 arXiv

  33. [41]

    Heterogeneous knowledge fusion: A novel approach for personalized recommendation via llm

    Bin Yin, Junjie Xie, Yu Qin, Zixiang Ding, Zhichao Feng, Xiang Li, and Wei Lin. Heterogeneous knowledge fusion: A novel approach for personalized recommendation via llm. In Proceedings of the 17th ACM Conference on Recommender Systems, pages 599–601, 2023

    2023

  34. [42]

    How would stance detection techniques evolve after the launch of chatgpt? arXiv preprint arXiv:2212.14548, 2022

    Bowen Zhang, Daijun Ding, Liwen Jing, Genan Dai, and Nan Yin. How would stance detection techniques evolve after the launch of chatgpt? arXiv preprint arXiv:2212.14548, 2022

    2022 arXiv

  35. [43]

    Benchmarking large language models for news summarization

    Tianyi Zhang, Faisal Ladhak, Esin Durmus, Percy Liang, Kathleen McKeown, and Tatsunori B Hashimoto. Benchmarking large language models for news summarization. Transactions of the Association for Computational Linguistics, 12:39–57, 2024

    2024

  36. [44]

    Generative job recommendations with large language model

    Zhi Zheng, Zhaopeng Qiu, Xiao Hu, Likang Wu, Hengshu Zhu, and Hui Xiong. Generative job recommendations with large language model. arXiv preprint arXiv:2307.02157, 2023

    2023 arXiv

  37. [45]

    Multilingual machine translation with large language models: Empirical results and analysis

    Wenhao Zhu, Hongyi Liu, Qingxiu Dong, Jingjing Xu, Shujian Huang, Lingpeng Kong, Jiajun Chen, and Lei Li. Multilingual machine translation with large language models: Empirical results and analysis. arXiv preprint arXiv:2304.04675, 2023

    2023 arXiv

  38. [46]

    "" Figure 3: The default system prompt used to generate social media posts across various personas. System prompt =

    Caleb Ziems, William Held, Omar Shaikh, Jiaao Chen, Zhehao Zhang, and Diyi Yang. Can large language models transform computational social science? Computational Linguistics, 50 (1):237–291, 2024. 13 A Prompts System prompt = """ You are a real American person who tweet about y...

    2024

  39. [2020]

    URL https://connect.apsanet.org/hsr/principles-and-guidance/

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.