# NepPlural: A Human-Anchored Benchmark for Persona Annotation in Nepali Migration Discourse

arXiv: https://arxiv.org/abs/2610.15732 · HTML: https://mlsanity.com/paper/2610.15732/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 38 references that are in the map.
- First posted on arXiv 11 Oct 2026.

## Builds on (most cited first, 23 of 23)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks](https://mlsanity.com/paper/2303.15056/) (2023)
- [Whose Opinions Do Language Models Reflect?](https://mlsanity.com/paper/2303.17548/) (2023)
- [Scalable agent alignment via reward modeling: a research direction](https://mlsanity.com/paper/1811.07871/) (2018)
- [Towards Measuring the Representation of Subjective Global Opinions in Language Models](https://mlsanity.com/paper/2306.16388/) (2023)
- [Assessing Cross-Cultural Alignment between ChatGPT and Human Societies: An Empirical Study](https://mlsanity.com/paper/2303.17466/) (2023)
- [Probing Pre-Trained Language Models for Cross-Cultural Differences in Values](https://mlsanity.com/paper/2203.13722/) (2022)
- [Investigating Cultural Alignment of Large Language Models](https://mlsanity.com/paper/2402.13231/) (2024)
- [Modular Pluralism: Pluralistic Alignment via Multi-LLM Collaboration](https://mlsanity.com/paper/2406.15951/) (2024)
- [The Ethics of Advanced AI Assistants](https://mlsanity.com/paper/2404.16244/) (2024)
- [PAD: Personalized Alignment of LLMs at Decoding-Time](https://mlsanity.com/paper/2410.04070/) (2024)
- [How Well Do LLMs Represent Values Across Cultures? Empirical Analysis of LLM Responses Based on Hofstede Cultural Dimensions](https://mlsanity.com/paper/2406.14805/) (2024)
- [Ethical Reasoning and Moral Value Alignment of LLMs Depend on the Language we Prompt them in](https://mlsanity.com/paper/2404.18460/) (2024)
- [Spectrum Tuning: Post-Training for Distributional Coverage and In-Context Steerability](https://mlsanity.com/paper/2510.06084/) (2025)
- [AI Alignment through Reinforcement Learning from Human Feedback? Contradictions and Limitations](https://mlsanity.com/paper/2406.18346/) (2024)
- [Does Cross-Cultural Alignment Change the Commonsense Morality of Language Models?](https://mlsanity.com/paper/2406.16316/) (2024)
- [Steerable Pluralism: Pluralistic Alignment via Few-Shot Comparative Regression](https://mlsanity.com/paper/2508.08509/) (2025)
- [Development of Pre-Trained Transformer-based Models for the Nepali Language](https://mlsanity.com/paper/2411.15734/) (2024)
- [Inference-Time Personalized Alignment with a Few User Preference Queries](https://mlsanity.com/paper/2511.02966/) (2025)
- [Histoires Morales: A French Dataset for Assessing Moral Alignment](https://mlsanity.com/paper/2501.17117/) (2025)
- [Consolidating and Developing Benchmarking Datasets for the Nepali Natural Language Understanding Tasks](https://mlsanity.com/paper/2411.19244/) (2024)

## Abstract (arXiv)

Pluralistic alignment asks that language models faithfully represent a society's diverse, competing values rather than collapse them into one dominant viewpoint, yet flagship pluralistic-alignment datasets remain heavily skewed toward Western respondents. We introduce NepPlural, a human-anchored benchmark for modelling competing socio-cultural personas in Nepali public discourse on youth migration and brain drain. We curate 2,148 code-mixed (Devanagari, Romanized Nepali, English) YouTube comments, annotate them along a four-axis persona taxonomy (Intent, Primary Driver, Value Orientation, Affect) plus an explicit No-Persona class, and build a 300-comment gold set labelled independently by three native Nepali speakers (mean Fleiss' kappa ≈ 0.55). Using this gold set to audit automatic annotation, we find that LLM annotation reliability is driven by protocol. Labels collected in batches through chat interfaces reach only near-chance inter-model agreement (kappa ≈ 0.17). Annotating one comment at a time raises agreement above the native-speaker level (kappa ≈ 0.72), but consistency is not validity: under a forced-choice schema, models agree on personas for the 39% of comments native speakers judge to express none, keeping macro-F1 against human gold near 0.50. Allowing abstention resolves this: models recover No-Persona comments with F1 of 0.79 to 0.87 and, against leave-one-annotator-out references, match or exceed individual native speakers (macro-F1 0.75 to 0.86 vs. 0.75). Encoders fine-tuned on the batched labels reach only 0.34. Finally, a human-judged steerability probe finds that three frontier models adopt single persona axes with 59 to 77% fidelity but compose full personas only 33% of the time, with no congruence gap detected at this scale between Western-congruent and Nepali-collectivist personas.