REVIEW 15 cited by
The PRISM Alignment Dataset: What Participatory, Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Human feedback is central to the alignment of Large Language Models (LLMs). However, open questions remain about methods (how), domains (where), people (who) and objectives (to what end) of feedback processes. To navigate these questions, we introduce PRISM, a dataset that maps the sociodemographics and stated preferences of 1,500 diverse participants from 75 countries, to their contextual preferences and fine-grained feedback in 8,011 live conversations with 21 LLMs. With PRISM, we contribute (i) wider geographic and demographic participation in feedback; (ii) census-representative samples for two countries (UK, US); and (iii) individualised ratings that link to detailed participant profiles, permitting personalisation and attribution of sample artefacts. We target subjective and multicultural perspectives on value-laden and controversial issues, where we expect interpersonal and cross-cultural disagreement. We use PRISM in three case studies to demonstrate the need for careful consideration of which humans provide what alignment data.
Forward citations
Cited by 15 Pith papers
-
Whose View of Safety? A Deep DIVE Dataset for Pluralistic Alignment of Text-to-Image Models
A demographically diverse annotation dataset shows that safety perceptions for text-to-image outputs vary by rater identity and that conventional safety classifiers under-detect bias harms flagged by minority-group raters.
-
Exposure is not manifestation: measurement target and output resolution jointly determine which behavioural-faithfulness evaluator wins
Small hyperbolic models (146M–3B) report 100% creative-seed preference, 90.7% compliance-gap detection, and a selective-gating skeleton–wallpaper memory pilot as a companion-AI stack.
-
CuMA: Aligning LLMs with Sparse Cultural Values via Demographic-Aware Mixture of Adapters
A demographic-conditioned mixture of LoRA experts improves LLM cultural alignment and reduces the tendency of dense models to produce generic, averaged responses.
-
The Alignment Veto: How Safety Training Suppresses Cultural Knowledge in LLMs
The full text builds the MENA Values benchmark (864 questions, 7 models) and reports that LLM cultural answers shift with language, decline with reasoning prompts, and hide strong internal preferences behind refusals—...
-
The Personalization Trap: How User Memory Alters Emotional Reasoning in LLMs
Adding user memory to LLMs degrades their emotional-intelligence test scores and systematically disadvantages marginalized user profiles.
-
CultureSynth: A Hierarchical Taxonomy-Guided and Retrieval-Augmented Framework for Cultural Question-Answer Synthesis
A taxonomy-guided retrieval-augmented framework generates CultureSynth-7, a multilingual cultural QA benchmark, and its evaluation of 14 LLMs suggests cultural competence emerges around 3B parameters.
-
Toward Socially Aware Vision-Language Models: Evaluating Cultural Competence Through Multimodal Story Generation
An evaluation of five VLMs on culturally-prompted multimodal story generation finds measurable cultural adaptation alongside metric bias and inverse alignment in some models.
-
From Seed to Harvest: Augmenting Human Creativity with AI for Red-teaming Text-to-Image Models
Seed2Harvest expands 1,000 human adversarial prompts into 27,650 LLM-generated variants that keep roughly comparable unsafe-image trigger rates and add hundreds of new geographic contexts.
-
CALMA: A Process for Deriving Context-aligned Axes for Language Model Alignment
CALMA is a grounded-theory, participatory method for deriving community-specific language model alignment axes from open-ended user interactions and group discussion, piloted with two small groups.
-
ModelCitizens: Representing Community Voices in Online Safety
A community-annotated toxicity dataset with conversational context shows that models trained on ingroup labels outperform state-of-the-art moderation APIs.
-
Amulet: Putting Complex Multi-Turn Conversations on the Stand with LLM Juries
Dialog-act and maxim-aware prompting improves LLM judge accuracy on multi-turn preference data by up to 8 points, with further gains from jury-style voting.
-
Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas
A new benchmark elicits AI models' value priorities from choices in 3,000 AI-risk dilemmas and reports correlations between those priorities and risky behaviors, including on the external HarmBench.
-
A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies
A taxonomy of 19 types of linguistic expressions and 5 guiding lenses for identifying when language technology outputs may contribute to anthropomorphism.
-
Can External Validation Tools Improve Annotation Quality for LLM-as-a-Judge?
Tool-augmented LLM annotators improve agreement with ground-truth preferences on long-form factual and coding tasks, with mixed results on math, compared to standard LLM-as-a-judge baselines.
-
Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges
A broad survey of LLM alignment that catalogs objectives, benchmarks, SFT/RLHF/DPO methods, and safety challenges, without contributing new experimental or theoretical results.
Discussion (0). Sign in to comment.