{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:U3XPIHHOPSJ5OBFKTT7AEN4Q3N","short_pith_number":"pith:U3XPIHHO","schema_version":"1.0","canonical_sha256":"a6eef41cee7c93d704aa9cfe023790db5fee0e4382756f58c06c952be5961d6a","source":{"kind":"arxiv","id":"2402.04379","version":2},"attestation_state":"computed","paper":{"title":"Fine-Tuned Language Models Generate Stable Inorganic Materials as Text","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cond-mat.mtrl-sci"],"primary_cat":"cs.LG","authors_text":"Andrea Madotto, Andrew Gordon Wilson, Anuroop Sriram, C. Lawrence Zitnick, Nate Gruver, Zachary Ulissi","submitted_at":"2024-02-06T20:35:28Z","abstract_excerpt":"We propose fine-tuning large language models for generation of stable materials. While unorthodox, fine-tuning large language models on text-encoded atomistic data is simple to implement yet reliable, with around 90% of sampled structures obeying physical constraints on atom positions and charges. Using energy above hull calculations from both learned ML potentials and gold-standard DFT calculations, we show that our strongest model (fine-tuned LLaMA-2 70B) can generate materials predicted to be metastable at about twice the rate (49% vs 28%) of CDVAE, a competing diffusion model. Because of t"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2402.04379","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-06T20:35:28Z","cross_cats_sorted":["cond-mat.mtrl-sci"],"title_canon_sha256":"b49516fd757773a339748090b876a60139d0aeb8494b0fea077468030514fe41","abstract_canon_sha256":"c6b53e8ea4c4b2cb8a0689e1d1b02cb8c305bc2ddd1b7a13f07822fa26ff9523"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:42:20.315137Z","signature_b64":"e+oPxvjxPPXmkuFTkMycF8proM54OVJxW0UXyPk4hR8etzAVHvABhA4KLYWolwoeHkZhzBEN7VL/eLYmtFzHAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a6eef41cee7c93d704aa9cfe023790db5fee0e4382756f58c06c952be5961d6a","last_reissued_at":"2026-07-05T11:42:20.314670Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:42:20.314670Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fine-Tuned Language Models Generate Stable Inorganic Materials as Text","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cond-mat.mtrl-sci"],"primary_cat":"cs.LG","authors_text":"Andrea Madotto, Andrew Gordon Wilson, Anuroop Sriram, C. Lawrence Zitnick, Nate Gruver, Zachary Ulissi","submitted_at":"2024-02-06T20:35:28Z","abstract_excerpt":"We propose fine-tuning large language models for generation of stable materials. While unorthodox, fine-tuning large language models on text-encoded atomistic data is simple to implement yet reliable, with around 90% of sampled structures obeying physical constraints on atom positions and charges. Using energy above hull calculations from both learned ML potentials and gold-standard DFT calculations, we show that our strongest model (fine-tuned LLaMA-2 70B) can generate materials predicted to be metastable at about twice the rate (49% vs 28%) of CDVAE, a competing diffusion model. Because of t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.04379","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2402.04379/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2402.04379","created_at":"2026-07-05T11:42:20.314724+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.04379v2","created_at":"2026-07-05T11:42:20.314724+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.04379","created_at":"2026-07-05T11:42:20.314724+00:00"},{"alias_kind":"pith_short_12","alias_value":"U3XPIHHOPSJ5","created_at":"2026-07-05T11:42:20.314724+00:00"},{"alias_kind":"pith_short_16","alias_value":"U3XPIHHOPSJ5OBFK","created_at":"2026-07-05T11:42:20.314724+00:00"},{"alias_kind":"pith_short_8","alias_value":"U3XPIHHO","created_at":"2026-07-05T11:42:20.314724+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21395","citing_title":"Atomistic Language Models Understand and Generate Materials","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20729","citing_title":"LLM-Guided Test-Time Discovery of Quantum-Chemical Approximation Algorithms","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31495","citing_title":"General-purpose LLMs as Constrained Crystal Composition Generators","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14759","citing_title":"Crys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinement","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14769","citing_title":"Composable Crystals: Controllable Materials Discovery via Concept Learning","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2507.16307","citing_title":"Perovskite-R1: a domain-specialized large language model for intelligent discovery of precursor additives and experimental design","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14344","citing_title":"CrystalReasoner: Reasoning and RL for Property-Conditioned Crystal Structure Generation","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03515","citing_title":"Scale-Dependent Input Representation and Confidence Estimation for LLMs in Materials Property Prediction","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27709","citing_title":"Conditional Generative Models Enable Targeted Exploration of MAX Phase Design Space","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27879","citing_title":"Generation of magnetic metal-organic frameworks","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03205","citing_title":"From Knowledge to Action: Outcomes of the 2025 Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry","ref_index":298,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11827","citing_title":"Inverse Design of Inorganic Compounds with Generative AI","ref_index":183,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N","json":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N.json","graph_json":"https://pith.science/api/pith-number/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/graph.json","events_json":"https://pith.science/api/pith-number/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/events.json","paper":"https://pith.science/paper/U3XPIHHO"},"agent_actions":{"view_html":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N","download_json":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N.json","view_paper":"https://pith.science/paper/U3XPIHHO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.04379&json=true","fetch_graph":"https://pith.science/api/pith-number/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/graph.json","fetch_events":"https://pith.science/api/pith-number/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/action/storage_attestation","attest_author":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/action/author_attestation","sign_citation":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/action/citation_signature","submit_replication":"https://pith.science/pith/U3XPIHHOPSJ5OBFKTT7AEN4Q3N/action/replication_record"}},"created_at":"2026-07-05T11:42:20.314724+00:00","updated_at":"2026-07-05T11:42:20.314724+00:00"}