{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6K3RCTXPJKISD4GRNUQ7MKNMHK","short_pith_number":"pith:6K3RCTXP","schema_version":"1.0","canonical_sha256":"f2b7114eef4a9121f0d16d21f629ac3a95f572ffed16fda9a1c469cdcf85bdf1","source":{"kind":"arxiv","id":"2405.09395","version":2},"attestation_state":"computed","paper":{"title":"Matching domain experts by training from scratch on domain knowledge","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"q-bio.NC","authors_text":"Bradley C. Love, Guangzhi Sun, Xiaoliang Luo","submitted_at":"2024-05-15T14:50:51Z","abstract_excerpt":"Recently, large language models (LLMs) have outperformed human experts in predicting the results of neuroscience experiments (Luo et al., 2024). What is the basis for this performance? One possibility is that statistical patterns in that specific scientific literature, as opposed to emergent reasoning abilities arising from broader training, underlie LLMs' performance. To evaluate this possibility, we trained (next word prediction) a relatively small 124M-parameter GPT-2 model on 1.3 billion tokens of domain-specific knowledge. Despite being orders of magnitude smaller than larger LLMs trained"},"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":"2405.09395","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"q-bio.NC","submitted_at":"2024-05-15T14:50:51Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"25100b19ece6374bdb785f3da15d0f0cf492306c2cf512ddc885e95912c9ab02","abstract_canon_sha256":"a01566b2b317f301d63968cfea136a54dcf1dc46847e006953d2f6915d78722c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:39:21.698504Z","signature_b64":"Xs6iFZCbJwp1BqpaGunqUypobvQ8hOOmLMAd2FV0EOvY6m4RX5Lt5Fm3dfRi/5py0HFtomyn+08gfNYuGZNHDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f2b7114eef4a9121f0d16d21f629ac3a95f572ffed16fda9a1c469cdcf85bdf1","last_reissued_at":"2026-07-05T08:39:21.698113Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:39:21.698113Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Matching domain experts by training from scratch on domain knowledge","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"q-bio.NC","authors_text":"Bradley C. Love, Guangzhi Sun, Xiaoliang Luo","submitted_at":"2024-05-15T14:50:51Z","abstract_excerpt":"Recently, large language models (LLMs) have outperformed human experts in predicting the results of neuroscience experiments (Luo et al., 2024). What is the basis for this performance? One possibility is that statistical patterns in that specific scientific literature, as opposed to emergent reasoning abilities arising from broader training, underlie LLMs' performance. To evaluate this possibility, we trained (next word prediction) a relatively small 124M-parameter GPT-2 model on 1.3 billion tokens of domain-specific knowledge. Despite being orders of magnitude smaller than larger LLMs trained"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.09395","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/2405.09395/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":"2405.09395","created_at":"2026-07-05T08:39:21.698166+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.09395v2","created_at":"2026-07-05T08:39:21.698166+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.09395","created_at":"2026-07-05T08:39:21.698166+00:00"},{"alias_kind":"pith_short_12","alias_value":"6K3RCTXPJKIS","created_at":"2026-07-05T08:39:21.698166+00:00"},{"alias_kind":"pith_short_16","alias_value":"6K3RCTXPJKISD4GR","created_at":"2026-07-05T08:39:21.698166+00:00"},{"alias_kind":"pith_short_8","alias_value":"6K3RCTXP","created_at":"2026-07-05T08:39:21.698166+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2411.11061","citing_title":"Beyond Human-Like Processing: Large Language Models Perform Equivalently on Forward and Backward Scientific Text","ref_index":37,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK","json":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK.json","graph_json":"https://pith.science/api/pith-number/6K3RCTXPJKISD4GRNUQ7MKNMHK/graph.json","events_json":"https://pith.science/api/pith-number/6K3RCTXPJKISD4GRNUQ7MKNMHK/events.json","paper":"https://pith.science/paper/6K3RCTXP"},"agent_actions":{"view_html":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK","download_json":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK.json","view_paper":"https://pith.science/paper/6K3RCTXP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.09395&json=true","fetch_graph":"https://pith.science/api/pith-number/6K3RCTXPJKISD4GRNUQ7MKNMHK/graph.json","fetch_events":"https://pith.science/api/pith-number/6K3RCTXPJKISD4GRNUQ7MKNMHK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK/action/storage_attestation","attest_author":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK/action/author_attestation","sign_citation":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK/action/citation_signature","submit_replication":"https://pith.science/pith/6K3RCTXPJKISD4GRNUQ7MKNMHK/action/replication_record"}},"created_at":"2026-07-05T08:39:21.698166+00:00","updated_at":"2026-07-05T08:39:21.698166+00:00"}