{"paper":{"title":"DVMap: Fine-Grained Pluralistic Value Alignment via High-Consensus Demographic-Value Mapping","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Mapping multi-dimensional demographics to values lets LLMs align with pluralistic preferences more tightly than national labels allow.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Deyi Xiong, Lei Yang, Pengyun Zhu, Yuqi Ren, Zhen Wang","submitted_at":"2026-05-14T06:06:43Z","abstract_excerpt":"Current Large Language Models (LLMs) typically rely on coarse-grained national labels for pluralistic value alignment. However, such macro-level supervision often obscures intra-country value heterogeneity, yielding a loose alignment. We argue that resolving this limitation requires shifting from national labels to multi-dimensional demographic constraints, which can identify groups with predictable, high-consensus value preference. To this end, we propose DVMap (High-Consensus Demographic-Value Mapping), a framework for fine-grained pluralistic value alignment. In this framework, we first pre"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experimental results demonstrate that DVMap effectively learns the manifold mapping from demographics to values, exhibiting strong generalization and robustness. On cross-demographic tests, Qwen3-8B-DVMap achieves 48.6% accuracy, surpassing the advanced open-source LLM DeepSeek-v3.2 (45.1%).","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That respondents with consistent value preferences under identical demographics form high-consensus groups whose preferences generalize reliably to unseen demographic combinations, countries, and value dimensions without introducing selection bias from the filtering process.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DVMap extracts high-consensus demographic groups from survey data and applies structured CoT plus GRPO to align LLMs with pluralistic values, reporting 48.6% accuracy on cross-demographic generalization tests.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Mapping multi-dimensional demographics to values lets LLMs align with pluralistic preferences more tightly than national labels allow.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e7efeb561748dd2928a5e4244546c1a37df70476bf280d3190bad63fd10bdac4"},"source":{"id":"2605.14420","kind":"arxiv","version":1},"verdict":{"id":"7b321340-df51-4e83-b4c7-f707353752d6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T02:00:43.241257Z","strongest_claim":"Experimental results demonstrate that DVMap effectively learns the manifold mapping from demographics to values, exhibiting strong generalization and robustness. On cross-demographic tests, Qwen3-8B-DVMap achieves 48.6% accuracy, surpassing the advanced open-source LLM DeepSeek-v3.2 (45.1%).","one_line_summary":"DVMap extracts high-consensus demographic groups from survey data and applies structured CoT plus GRPO to align LLMs with pluralistic values, reporting 48.6% accuracy on cross-demographic generalization tests.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That respondents with consistent value preferences under identical demographics form high-consensus groups whose preferences generalize reliably to unseen demographic combinations, countries, and value dimensions without introducing selection bias from the filtering process.","pith_extraction_headline":"Mapping multi-dimensional demographics to values lets LLMs align with pluralistic preferences more tightly than national labels allow."},"references":{"count":13,"sample":[{"doi":"","year":2024,"title":"Association for Computational Linguistics","work_id":"0f9b87ad-f0c9-436f-9b1b-faa8e4a7c8ac","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"In9th International Conference on Learning Representa- tions, ICLR 2021, Virtual Event, Austria, May 3-7,","work_id":"c1e314ad-45cb-40dd-b862-9ec0b81988ab","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2005,"title":"The ghost in the machine has an american accent: value conflict in gpt-3","work_id":"9324c45e-22da-43e3-ab89-872beb2e2d38","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","ref_index":4,"cited_arxiv_id":"2410.21276","is_internal_anchor":true},{"doi":"","year":2023,"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","ref_index":5,"cited_arxiv_id":"2402.03300","is_internal_anchor":true}],"resolved_work":13,"snapshot_sha256":"263acef597a6a6c4e7132dd426fba196803313f84736e1e0457180bce9ceca5d","internal_anchors":3},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}