Pith. sign in

REVIEW 3 major objections 6 minor 50 references

Prompt programming with a compilable prompt optimizer moves open-weight LLMs closer to target cultures on the Inglehart–Welzel map, often beating a hand-written 'you are a citizen of X' prefix.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 17:58 UTC pith:D22TYTQS

load-bearing objection Useful combination of DSPy and survey-grounded cultural distance, but the headline claim that optimization beats manual prompting lacks the reported numbers to back it up. the 3 major comments →

arxiv 2603.16827 v2 pith:D22TYTQS submitted 2026-03-17 cs.AI cs.CL

Prompt Programming for Cultural Bias and Alignment of Large Language Models

Maksim Eren , Eric Michalak , Brian Cook , Johnny Seales Jr This is my paper
classification cs.AI cs.CL
keywords LLMcultural biascultural alignmentprompt engineeringprompt programmingDSPyInglehart–Welzel mapopen-weight models
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tries to establish that a hand-written cultural prompt—'You are a citizen of X'—is not the best way to align an LLM with a target country's values. It shows that treating the culture instruction as an optimizable text parameter, compiled by a DSPy-style prompt optimizer against a cultural-distance objective, moves open-weight models closer to human survey benchmarks than the manual prefix for most model–country pairs. The best configuration, MIPROv2 with a large instruction-proposal model, reduces cultural distance beyond manual prompting for every model except one. The authors first reproduce the existing survey-based framework on five open-weight models, confirming that generic prompting clusters all models near Western value profiles. If the claim holds, prompt compilation offers a more stable, transferable route to culturally aligned responses than manual template design.

Core claim

On the paper's own terms, the central discovery is that cultural conditioning can be posed as an optimization problem and solved by prompt programming: instead of fixing a single persona template, the instruction is treated as a discrete text parameter and tuned, per target model and country set, to minimize Euclidean distance in the Inglehart–Welzel cultural map. Generic prompting places all five open-weight models in a compact Western-skewed region; manual country-identity prompting shifts them toward the target benchmark for many countries; DSPy compilation with MIPROv2 using a 120B proposal model achieves the largest additional reductions beyond manual prompting for four of five models.

What carries the argument

The cultural-distance objective: LLM answers to ten Integrated Values Surveys items are projected into the two-dimensional Inglehart–Welzel space (Survival vs. Self-Expression; Traditional vs. Secular) using a varimax-rotated PCA fitted on human survey data, and alignment is scored as Euclidean distance to the human country benchmark. DSPy's MIPROv2 teleprompter, which proposes candidate instructions with a language model and selects the best combination by Bayesian optimization over the discrete instruction space, is what carries the argument: it is the mechanism that produces the improved prompts.

Load-bearing premise

The whole comparison rests on treating Euclidean distance in the Inglehart–Welzel projection built from ten forced-choice survey items as a valid measure of cultural alignment; if that map misses how values actually surface in real language use, the reported improvements measure optimization against a survey proxy.

What would settle it

Take the best DSPy-compiled prompt for a non-Western country and generate open-ended advice on a policy dilemma; ask native raters from that country to judge whether the compiled-prompt output is more culturally appropriate than the manual-prefix output. If raters do not prefer the compiled-prompt outputs despite its lower Inglehart–Welzel distance, the distance metric is not measuring cultural alignment.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Open-weight LLMs share a Western-skewed default prior under generic prompting; default outputs will misalign with non-Western target populations in policy or document-engineering tasks.
  • Manual country-identity prompting reduces cultural distance for many countries, confirming that explicit persona framing is a usable and cheap alignment lever.
  • DSPy-style compilation can reduce distance further, but only when paired with a capable proposer (MIPROv2 + 120B model), so optimizer choice matters in practice.
  • Alignment is incomplete: some countries and territories remain outliers after optimization, so evaluation should be country-disaggregated rather than averaged globally.
  • A compiled prompt is model- and training-set-specific; the paper's cross-validation over countries provides a template for measuring transfer.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A natural extension is testing whether the compiled prompts reduce distance on open-ended generation and multi-turn dialogue, not just forced-choice survey items; the paper itself flags this gap.
  • If prompt text is portable, the same compiled instructions might be transferred across models within a family or to new countries; a direct test would compare distance reductions on held-out countries when using a prompt compiled on one model versus another.
  • The improvements likely operate by moving responses along the Survival vs. Self-Expression axis; extracting the lexical content of compiled prompts could reveal a general alignment strategy that could be applied without optimization.
  • The optimization-by-distance approach could be re-run against other cultural maps (e.g., Hofstede or moral-foundation dimensions) to see whether the gains reflect true value alignment or are an artifact of the specific two-dimensional projection.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper replicates Tao et al.'s survey-grounded cultural alignment framework on five open-weight LLMs and proposes using DSPy prompt programming to optimize cultural conditioning against an IVS-based cultural-distance objective. It reports that generic prompting yields a Western-skewed default profile, manual country prompting reduces distance, and DSPy-compiled prompts, especially MIPROv2 with GPT-OSS:120B as proposer, often improve further. The central empirical claim is that prompt optimization beats manual prompt engineering, supported by a 5-fold country-level cross-validation design, but the results section presents only qualitative descriptions and figures with no reported numeric distances, effect sizes, or significance tests.

Significance. If the empirical claims hold, the contribution is useful: it extends a prominent cultural-alignment benchmark to open-weight models, introduces a principled prompt-optimization baseline, and evaluates generalization across held-out countries. The external IVS benchmark and PCA projection are not invented by the authors, the 5-fold cross-validation over countries is a sound design choice, and the distinction between proposer and target model is clearly described. These are genuine strengths. However, the manuscript's central claim—that DSPy prompt programming often outperforms manual prompt engineering—is not verifiable from the reported text, which severely limits the current contribution's evidentiary value.

major comments (3)
  1. [Section 4, Eqs. (9)-(10), (18)] The abstract and Section 4 state that 'prompt optimization often improves upon cultural prompt engineering' and that 'MIPROv2 with GPT-OSS:120B yields the largest additional reductions beyond manual prompting for every model except Llama 3.3.' Yet the paper never reports the numeric values of d_man, d_DSPy, Δd_man, Δd_DSPy, the fraction of countries with Δd<0, or the held-out d_test from Eq. (18). No confidence intervals, standard errors, or paired significance tests are given. Figure 2 is a qualitative visualization; without the underlying numbers, a reader cannot assess whether the advantage is real, statistically reliable, or concentrated in a few countries. This is the load-bearing evidence for the central claim and must be provided.
  2. [Section 3.4, Eq. (18)] The 5-fold country-level cross-validation is a strength, but the held-out metric d_test is never disclosed. The only quantitative values in the paper are the per-country deltas in Figure 3 for a single configuration (gpt-oss:120b with MIPROv2). Without d_test or per-fold summaries for all five models and all six DSPy configurations, the generalization claim cannot be verified. The authors should report a table of d_test for each model and condition, plus the fraction of countries improved, with variability across folds.
  3. [Section 4, Figure 2] The qualitative statement 'for Gemma 3 and the GPT-OSS target models, the gains are more selective: only MIPROv2 with the GPT-OSS:120B proposer consistently outperforms manual prompting' is not supported by any reported statistic. 'Consistently outperforms' implies a comparison across countries, but no per-country agreement rates, win/tie/loss counts, or effect sizes are given. The authors should quantify the comparison, e.g., with the distribution of Δd_DSPy vs. Δd_man across countries and a paired test.
minor comments (6)
  1. [Discussion, Limitations] The paper acknowledges that the forced-choice survey instrument 'may not reflect how cultural values surface in open-ended generation, multi-turn dialogue, or decision-support deployments.' This is an important caveat and should be echoed in the abstract or introduction so readers do not over-interpret the distance metric as a complete measure of cultural alignment.
  2. [Section 3.4] The manual baseline is only the minimal 'You are a citizen of X' prefix. Since the paper argues for the value of prompt programming over 'manual prompt engineering,' it would be helpful to specify whether any manual prompt engineering beyond the fixed prefix was attempted, or whether the baseline is intentionally minimal. If the latter, state this explicitly.
  3. [Figure 1] The colored areas around each point are described as 'visual purposes of highlighting clustering' but the figure caption does not explain how the area is computed (e.g., convex hull, kernel density). Please clarify or remove if purely decorative.
  4. [Section 3.2, Eq. (5)] The persona-variant averaging is described but the number and exact nature of the variants are not given beyond examples. A short list or reference to a table would improve reproducibility.
  5. [General] The paper uses 'open-weight' and 'open-source' interchangeably; since the models are open-weight but not necessarily open-source in all cases, pick one term and use it consistently.
  6. [References] DSPy-related references [7,8] are to documentation pages rather than archival papers or stable releases. If possible, cite the versioned documentation or the primary DSPy paper for the teleprompter algorithms.

Circularity Check

0 steps flagged

No significant circularity: the DSPy comparison is a held-out prediction against an external IVS benchmark.

full rationale

The paper's derivation chain is self-contained with respect to circularity concerns. The cultural map, country coordinates, and projection weights come from external sources (IVS survey data and Tao et al. [42]), not from the models or prompts under evaluation. The DSPy optimization minimizes a cultural-distance objective (Eqs. 11-14) that is the same metric later used for evaluation (Eqs. 8, 18), but this identity is broken by the specified 5-fold cross-validation protocol: compilation is performed only on the training pool, and the reported mean held-out distance d_test is computed on held-out countries. Thus, the central claim that prompt optimization improves over manual prompting is a genuine generalization test rather than a fitted-input-called-prediction. The only self-citation in the paper (Eren & Perez [9]) appears in the related-work discussion and is not load-bearing for any result. The acknowledged limitation that the IVS forced-choice instrument may not capture cultural alignment in open-ended deployments is a construct-validity issue affecting all compared regimes equally, not a circularity. The absence of numeric results, confidence intervals, and significance tests in Section 4 is a reporting/evidence concern, not a demonstration that the result reduces to its inputs. No specific equation or quoted passage exhibits a circular reduction, so per the hard rules no circularity is claimed.

Axiom & Free-Parameter Ledger

2 free parameters · 6 axioms · 0 invented entities

The paper introduces no new theoretical entities and fits no numeric constants of its own. Its central comparisons inherit the IVS projection weights and rescaling from Tao et al., so the benchmark coordinate system is treated as given. The main tuned quantity is the discrete prompt instruction, optimized on training countries and evaluated on held-out countries.

free parameters (2)
  • IVS rescaling constants = PC1: 1.81, -0.38 offset; PC2: 1.61, -0.01 offset
    From Tao et al. Eqs. (1-2), used to map PCA scores to canonical Inglehart–Welzel axes. Inherited unchanged; all distances are computed in this rescaled space.
  • IVS PCA/varimax scoring weights = W_rot (2x10), not listed in text
    Estimated from IVS respondent data in prior work and reused in Eq. (4). The entire projection of model responses depends on these weights, but they are not re-fit in this paper.
axioms (6)
  • domain assumption Ten IVS/WVS/EVS forced-choice items and the Inglehart–Welzel axes validly operationalize culture for alignment purposes.
    Invoked in Section 3.1, where the benchmark space is constructed. The authors note in the Discussion that this is a limitation for open-ended behavior.
  • domain assumption A fixed persona prefix such as 'You are a citizen of X' is a valid cultural-conditioning mechanism for LLMs.
    Used in Section 3.2 as the manual prompting baseline; inherited from Tao et al.
  • domain assumption Euclidean distance in the two-dimensional IVS map measures cultural misalignment.
    Stated in Section 3.3 as the definition of country-level cultural distance; the paper's entire objective is to minimize this distance.
  • domain assumption The linear rescaling of Tao et al. applies to open-weight model projections.
    Applied in Eqs. (1-2) to all model projections; the paper does not re-estimate these constants.
  • domain assumption DSPy teleprompters (COPRO and MIPROv2) perform a genuine optimization over prompt instructions.
    Relied on in Section 3.4; the paper does not analyze how the candidate search interacts with target-model stochasticity.
  • standard math Principal component analysis with varimax rotation preserves the canonical two-dimensional structure of the IVS data.
    Standard statistical method used in Section 3.1, following prior cultural-map literature.

pith-pipeline@v1.3.0-alltime-deepseek · 15909 in / 13154 out tokens · 133846 ms · 2026-08-02T17:58:12.659600+00:00 · methodology

0 comments
read the original abstract

Culture shapes reasoning, values, prioritization, and strategic decision-making, yet large language models (LLMs) often exhibit cultural biases that misalign with target populations. As LLMs are increasingly used for strategic decision-making, policy support, and document engineering tasks such as summarization, categorization, and compliance-oriented auditing, improving cultural alignment is important for ensuring that downstream analyses and recommendations reflect target-population value profiles rather than default model priors. Previous work introduced a survey-grounded cultural alignment framework and showed that culture-specific prompting can reduce misalignment, but it primarily evaluated proprietary models and relied on manual prompt engineering. In this paper, we validate and extend that framework by reproducing its social sciences survey based projection and distance metrics on open-weight LLMs, testing whether the same cultural skew and benefits of culture conditioning persist outside closed LLM systems. Building on this foundation, we introduce use of prompt programming with DSPy for this problem-treating prompts as modular, optimizable programs-to systematically tune cultural conditioning by optimizing against cultural-distance objectives. In our experiments, we show that prompt optimization often improves upon cultural prompt engineering, suggesting prompt compilation with DSPy can provide a more stable and transferable route to culturally aligned LLM responses.

Figures

Figures reproduced from arXiv: 2603.16827 by Brian Cook, Eric Michalak, Johnny Seales Jr, Maksim Eren.

Figure 1
Figure 1. Figure 1: Cultural map of countries/territories in the IVS benchmark space (Survival vs. Self-Expression; Traditional vs. Secular [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Country-level cultural distance for open-source LLMs under three prompting regimes: (i) without culture conditioning, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Per-country movement in the Inglehart–Welzel IVS benchmark space (PC1 [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 1 canonical work pages

  1. [1]

    ElNokrashy, Mai Alkhamissi, and Mona T

    Badr AlKhamissi, Muhammad N. ElNokrashy, Mai Alkhamissi, and Mona T. Diab. 2024. Investigating Cultural Alignment of Large Language Models.ArXiv abs/2402.13231 (2024). https://api.semanticscholar.org/CorpusID:267759574

    Pith/arXiv arXiv 2024

  2. [2]

    Arnav Arora, Lucie-Aimée Kaffee, and Isabelle Augenstein. 2022. Probing Pre- Trained Language Models for Cross-Cultural Differences in Values.ArXiv abs/2203.13722 (2022). https://api.semanticscholar.org/CorpusID:247748753

    Pith/arXiv arXiv 2022

  3. [3]

    Xue, Peter S

    Mohammad Atari, Mona J. Xue, Peter S. Park, Damián E. Blasi, and Joseph Henrich. 2023. Which Humans?PsyArXiv(2023). doi:10.31234/osf.io/5b26t

    doi:10.31234/osf.io/5b26t 2023

  4. [4]

    Miller, Sandra Mitts, Adithya Renduchintala, Stephen Roller, Dirk Rowe, Weiyan Shi, Joe Spisak, Alexander Wei, David J

    Anton Bakhtin, Noam Brown, Emily Dinan, Gabriele Farina, Colin Flaherty, Daniel Fried, Andrew Goff, Jonathan Gray, Hengyuan Hu, Athul Paul Jacob, Mo jtaba Komeili, Karthik Konath, Minae Kwon, Adam Lerer, Mike Lewis, Alexan- der H. Miller, Sandra Mitts, Adithya Renduchintala, Stephen Roller, Dirk Rowe, Weiyan Shi, Joe Spisak, Alexander Wei, David J. Wu, Hu...

    2022

  5. [5]

    Michal Bravansky, Filip Trhlík, and Fazl Barez. 2025. Rethinking AI Cultural Alignment. https://api.semanticscholar.org/CorpusID:275515905

    2025

  6. [6]

    Bram Bult’e and Ayla Rigouts Terryn. 2025. LLMs and Cultural Values: the Impact of Prompt Language and Explicit Cultural Framing.ArXivabs/2511.03980 (2025). https://api.semanticscholar.org/CorpusID:282812238

    arXiv 2025

  7. [7]

    DSPy Contributors. 2024. DSPy Documentation: COPRO Teleprompter. https: //dspy.ai/api/teleprompt/COPRO/. Accessed: 2026-02-24

    2024

  8. [8]

    DSPy Contributors. 2024. DSPy Documentation: MIPROv2 Optimizer. https: //dspy.ai/api/optimizers/MIPROv2/. Accessed: 2026-02-24

    2024

  9. [9]

    Maksim E Eren and Dorianis M Perez. 2025. Rethinking Science in the Age of Artificial Intelligence.arXiv preprint arXiv:2511.10524(2025)

    arXiv 2025

  10. [10]

    European Values Study. 2022. European Values Study 2017–2022: Trend File. https://europeanvaluesstudy.eu/. Accessed: 2026-02-24

    2022

  11. [11]

    Kimo Gandall, Juliana Chhouk, Alex Wang, and Logan Knight. 2022. Predicting Policy: A Psycholinguistic Artificial Intelligence in the United Nations.Social Science Computer Review41 (2022), 410 – 437. https://api.semanticscholar.org/ CorpusID:248896975

    2022

  12. [12]

    Gemma Team. 2024. Gemma: Open Models Based on Gemini Research and Technology. https://arxiv.org/abs/2403.08295. Accessed 2026-02-24

    Pith/arXiv arXiv 2024

  13. [13]

    Aditi Godbole, Jabin Geevarghese George, and Smita Shandilya. 2024. Leveraging Long-Context Large Language Models for Multi-Document Understanding and Summarization in Enterprise Applications.ArXivabs/2409.18454 (2024). https: //api.semanticscholar.org/CorpusID:272969413

    Pith/arXiv arXiv 2024

  14. [14]

    Candida Maria Greco, Lucio La Cava, and Andrea Tagarelli. 2026. Culturally Grounded Personas in Large Language Models: Characterization and Alignment with Socio-Psychological Value Frameworks.ArXivabs/2601.22396 (2026). https: //api.semanticscholar.org/CorpusID:285241197

    Pith/arXiv arXiv 2026

  15. [15]

    Mourad Gridach, Jay Nanavati, Khaldoun Zine El Abidine, Lenon Mendes, and Christina DeFilippo Mack. 2025. Agentic AI for Scientific Discovery: A Survey of Progress, Challenges, and Future Directions.ArXivabs/2503.08979 (2025). https://api.semanticscholar.org/CorpusID:276937648

    Pith/arXiv arXiv 2025

  16. [16]

    Christian Härpfer, Ronald Inglehart, et al. 2022. World Values Survey: Round Seven — Country-Pooled Datafile (2017–2022), Version 5.0. https://www. worldvaluessurvey.org/WVSDocumentationWV7.jsp. Accessed: 2026-02-24

    2022

  17. [17]

    Geert Hofstede. 2001. Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations Across Nations. https://api.semanticscholar.org/ CorpusID:145428141

    2001

  18. [18]

    Hogan and Andrea Brennen

    Daniel P. Hogan and Andrea Brennen. 2024. Open-Ended Wargames with Large Language Models.ArXivabs/2404.11446 (2024). https://api.semanticscholar.org/ CorpusID:269187731

    Pith/arXiv arXiv 2024

  19. [19]

    2005.Modernization, Cultural Change, and Democracy: The Human Development Sequence

    Ronald Inglehart and Christian Welzel. 2005.Modernization, Cultural Change, and Democracy: The Human Development Sequence. Cambridge University Press

    2005

  20. [20]

    Rebecca L Johnson, Giada Pistilli, Natalia Menédez-González, Leslye Denisse Dias Duran, Enrico Panai, Julija Kalpokiene, and Donald Jay Bertulfo. 2022. The Ghost in the Machine has an American accent: value conflict in GPT-3.arXiv preprint arXiv:2203.07785(2022)

    Pith/arXiv arXiv 2022

  21. [21]

    Alastair Iain Johnston. 1995. Thinking about strategic culture.International security19, 4 (1995), 32–64. Eren and Michalak et al

    1995

  22. [22]

    I. T. Jolliffe and J. Cadima. 2016.Principal Component Analysis(2 ed.). Springer. doi:10.1007/978-1-4757-1904-8

    doi:10.1007/978-1-4757-1904-8 2016

  23. [23]

    Henry F. Kaiser. 1958. The Varimax Criterion for Analytic Rotation in Factor Analysis.Psychometrika23, 3 (1958), 187–200. doi:10.1007/BF02289233

    doi:10.1007/bf02289233 1958

  24. [24]

    Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, et al. 2023. DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines.arXiv preprint arXiv:2310.03714(2023). https://arxiv.org/abs/2310.03714

    Pith/arXiv arXiv 2023

  25. [25]

    Lawrence Kuznar, Nicole Heath, and George Popp. 2023. Strategic culture— Its history, issues, and complexity. (2023). Prepared for Strategic Multilayer Assessment, Joint Staff, J3

    2023

  26. [26]

    Louis Kwok, Michal Bravansky, and Lewis Griffin. 2024. Evaluating Cultural Adaptability of a Large Language Model via Simulation of Synthetic Personas. InFirst Conference on Language Modeling. https://openreview.net/forum?id= S4ZOkV1AHl

    2024

  27. [27]

    Florian Königstorfer and Stefan Thalmann. 2022. AI Documentation: A path to accountability.Journal of Responsible Technology11 (2022), 100043. doi:10.1016/ j.jrt.2022.100043

    arXiv 2022

  28. [28]

    Jiya Manchanda, Laura Boettcher, Matheus Westphalen, and Jasser Jasser

  29. [29]

    Meta. 2024. Llama 3 Model Card. https://github.com/meta-llama/llama3/blob/ main/MODEL_CARD.md. Accessed 2026-02-24

    2024

  30. [30]

    Meta. 2025. Llama 4 Model Card. https://github.com/meta-llama/llama-models/ blob/main/models/llama4/MODEL_CARD.md. Accessed 2026-02-24

    2025

  31. [31]

    Tarek Naous, Michael Joseph Ryan, and Wei Xu. 2023. Having Beer after Prayer? Measuring Cultural Bias in Large Language Models. InAnnual Meeting of the Asso- ciation for Computational Linguistics. https://api.semanticscholar.org/CorpusID: 258865272

    2023

  32. [32]

    Roberto Navigli, Simone Conia, and Björn Ross. 2023. Biases in Large Language Models: Origins, Inventory, and Discussion.J. Data and Information Quality15, 2, Article 10 (June 2023), 21 pages. doi:10.1145/3597307

    doi:10.1145/3597307 2023

  33. [33]

    Juhyun Oh, Inha Cha, Michael Saxon, Hyunseung Lim, Shaily Bhatt, and Alice Oh

  34. [34]

    Alexandra O’Neil, Daniel Swanson, and Shobhana Chelliah. 2024. Computational Language Documentation: Designing a Modular Annotation and Data Manage- ment Tool for Cross-cultural Applicability. InProceedings of the 2nd Workshop on Cross-Cultural Considerations in NLP, Vinodkumar Prabhakaran, Sunipa Dev, Luciana Benotti, Daniel Hershcovich, Laura Cabello, Y...

    work page doi:10.18653/v1/2024.c3nlp-1.9 2024

  35. [35]

    OpenAI. 2025. gpt-oss-120b & gpt-oss-20b Model Card. https://arxiv.org/abs/ 2508.10925. Accessed 2026-02-24

    Pith/arXiv arXiv 2025

  36. [36]

    Siddhesh Milind Pawar, Junyeong Park, Jiho Jin, Arnav Arora, Junho Myung, Srishti Yadav, Faiz Ghifari Haznitrama, Inhwa Song, Alice Oh, and Isabelle Augenstein. 2024. Survey of Cultural Awareness in Language Models: Text and Beyond.ArXivabs/2411.00860 (2024). https://api.semanticscholar.org/CorpusID: 273811670

    Pith/arXiv arXiv 2024

  37. [37]

    Kenneth Payne. 2026. AI Arms and Influence: Frontier Models Exhibit Sophisti- cated Reasoning in Simulated Nuclear Crises.arXiv preprint arXiv:2602.14740 (2026)

    arXiv 2026

  38. [38]

    Kroiz, Feileen Li, Hudson Tao, Ashay Srivastava, Hevander Da Costa, Saloni Gupta, Megan L

    Sander Schulhoff, Michael Ilie, Nishant Balepur, Konstantine Kahadze, Amanda Liu, Chenglei Si, Yinheng Li, Aayush Gupta, HyoJung Han, Sevien Schulhoff, Pranav Sandeep Dulepet, Saurav Vidyadhara, Dayeon Ki, Sweta Agrawal, Chau Pham, Gerson C. Kroiz, Feileen Li, Hudson Tao, Ashay Srivastava, Hevander Da Costa, Saloni Gupta, Megan L. Rogers, Inna Goncearenco...

    2024

  39. [39]

    Elina Sigdel and Anastasia Panfilova. 2026. RusLICA: A Russian-Language Platform for Automated Linguistic Inquiry and Category Analysis.arXiv preprint arXiv:2601.20275(2026)

    arXiv 2026

  40. [40]

    Jack Snyder. 1977. The Soviet Strategic Culture. Implications for Limited Nuclear Operations. https://api.semanticscholar.org/CorpusID:150401111

    1977

  41. [41]

    Julius Steen and Katja Markert. 2023. Bias in News Summarization: Measures, Pitfalls and Corpora. InAnnual Meeting of the Association for Computational Linguistics. https://api.semanticscholar.org/CorpusID:262013727

    2023

  42. [42]

    Yan Tao, Olga Viberg, Ryan S Baker, and René F Kizilcec. 2024. Cul- tural bias and cultural alignment of large language models.PNAS Nexus 3, 9 (09 2024), pgae346. arXiv:https://academic.oup.com/pnasnexus/article- pdf/3/9/pgae346/59151559/pgae346.pdf doi:10.1093/pnasnexus/pgae346

    doi:10.1093/pnasnexus/pgae346 2024

  43. [43]

    Jonathan Tappe and Fredrik Doeser. 2021. A machine learning approach to the study of German strategic culture.Contemporary Security Policy42, 4 (2021), 450–474

    2021

  44. [44]

    Mustafa Tuna, Kristina Schaaff, and Tim Schlippe. 2024. Effects of Language- and Culture-Specific Prompting on ChatGPT. In2024 2nd International Conference on Foundation and Large Language Models (FLLM). 73–81. doi:10.1109/FLLM63129. 2024.10852463

    arXiv 2024

  45. [45]

    World Values Survey Association and European Values Study. 2023. Inte- grated Values Surveys (IVS) — Codebook and Documentation. https://www. worldvaluessurvey.org/WVSDocumentationWVL.jsp. Accessed: 2026-02-24

    2023

  46. [46]

    Xufeng Yao, Xiaoxu Wu, Xi Li, Huan Xu, Chenlei Li, Ping Huang, Sitong Li, Xiaoning Ma, and Jiulong Shan. 2024. Smart Audit System Empowered by LLM.ArXivabs/2410.07677 (2024). https://api.semanticscholar.org/CorpusID: 273233253

    Pith/arXiv arXiv 2024

  47. [47]

    Wenlong Zhao, Debanjan Mondal, Niket Tandon, Danica Dillion, Kurt Gray, and Yu Gu. 2024. WorldValuesBench: A Large-Scale Benchmark Dataset for Multi- Cultural Value Awareness of Language Models. InInternational Conference on Language Resources and Evaluation. https://api.semanticscholar.org/CorpusID: 269362884

    2024

  48. [48]

    Ke Zhou, Marios Constantinides, and Daniele Quercia. 2025. Should LLMs be WEIRD? Exploring WEIRDness and Human Rights in Large Language Mod- els.ArXivabs/2508.19269 (2025). https://api.semanticscholar.org/CorpusID: 280919174

    Pith/arXiv arXiv 2025

  49. [2024]

    https://api.semanticscholar.org/CorpusID:274788875

    The Open Source Advantage in Large Language Models (LLMs).ArXiv abs/2412.12004 (2024). https://api.semanticscholar.org/CorpusID:274788875

    arXiv 2024

  50. [2025]

    https://api.semanticscholar.org/CorpusID:281079526

    Culture is Everywhere: A Call for Intentionally Cultural Evaluation.ArXiv abs/2509.01301 (2025). https://api.semanticscholar.org/CorpusID:281079526

    arXiv 2025