{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:LUS63OLPNFXPQTK4SMY27NKYWC","short_pith_number":"pith:LUS63OLP","schema_version":"1.0","canonical_sha256":"5d25edb96f696ef84d5c9331afb558b0a487ff7af6ac67f788dcc444a9a367de","source":{"kind":"arxiv","id":"2310.04680","version":1},"attestation_state":"computed","paper":{"title":"The Cost of Down-Scaling Language Models: Fact Recall Deteriorates before In-Context Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Gintare Karolina Dziugaite, Jonathan Ragan-Kelley, Michael Carbin, Nolan Clement, Tian Jin, Vaishnavh Nagarajan, Xin Dong","submitted_at":"2023-10-07T03:36:39Z","abstract_excerpt":"How does scaling the number of parameters in large language models (LLMs) affect their core capabilities? We study two natural scaling techniques -- weight pruning and simply training a smaller or larger model, which we refer to as dense scaling -- and their effects on two core capabilities of LLMs: (a) recalling facts presented during pre-training and (b) processing information presented in-context during inference. By curating a suite of tasks that help disentangle these two capabilities, we find a striking difference in how these two abilities evolve due to scaling. Reducing the model size "},"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":"2310.04680","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-10-07T03:36:39Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"536fa4a41dd3cfeacbdeb9ba01df8ca4cbceac6734a9a0bd56c7a1c4ec2a3158","abstract_canon_sha256":"c3a006dec9081fb300820d89f645ada8d0667e61ac6a1ddcc52e25032ec1bbff"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T01:09:10.839380Z","signature_b64":"c7+FZr1wkJU9XoDCmjhBTXxGnnQfzj7zeVccN0UBqYvORVIR1rhfnd6u1Oo3jvZEWNCD3GfWcVjjCPe3ayowBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5d25edb96f696ef84d5c9331afb558b0a487ff7af6ac67f788dcc444a9a367de","last_reissued_at":"2026-06-10T01:09:10.838380Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T01:09:10.838380Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Cost of Down-Scaling Language Models: Fact Recall Deteriorates before In-Context Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Gintare Karolina Dziugaite, Jonathan Ragan-Kelley, Michael Carbin, Nolan Clement, Tian Jin, Vaishnavh Nagarajan, Xin Dong","submitted_at":"2023-10-07T03:36:39Z","abstract_excerpt":"How does scaling the number of parameters in large language models (LLMs) affect their core capabilities? We study two natural scaling techniques -- weight pruning and simply training a smaller or larger model, which we refer to as dense scaling -- and their effects on two core capabilities of LLMs: (a) recalling facts presented during pre-training and (b) processing information presented in-context during inference. By curating a suite of tasks that help disentangle these two capabilities, we find a striking difference in how these two abilities evolve due to scaling. Reducing the model size "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.04680","kind":"arxiv","version":1},"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/2310.04680/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":"2310.04680","created_at":"2026-06-10T01:09:10.838516+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.04680v1","created_at":"2026-06-10T01:09:10.838516+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.04680","created_at":"2026-06-10T01:09:10.838516+00:00"},{"alias_kind":"pith_short_12","alias_value":"LUS63OLPNFXP","created_at":"2026-06-10T01:09:10.838516+00:00"},{"alias_kind":"pith_short_16","alias_value":"LUS63OLPNFXPQTK4","created_at":"2026-06-10T01:09:10.838516+00:00"},{"alias_kind":"pith_short_8","alias_value":"LUS63OLP","created_at":"2026-06-10T01:09:10.838516+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC","json":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC.json","graph_json":"https://pith.science/api/pith-number/LUS63OLPNFXPQTK4SMY27NKYWC/graph.json","events_json":"https://pith.science/api/pith-number/LUS63OLPNFXPQTK4SMY27NKYWC/events.json","paper":"https://pith.science/paper/LUS63OLP"},"agent_actions":{"view_html":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC","download_json":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC.json","view_paper":"https://pith.science/paper/LUS63OLP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.04680&json=true","fetch_graph":"https://pith.science/api/pith-number/LUS63OLPNFXPQTK4SMY27NKYWC/graph.json","fetch_events":"https://pith.science/api/pith-number/LUS63OLPNFXPQTK4SMY27NKYWC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC/action/storage_attestation","attest_author":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC/action/author_attestation","sign_citation":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC/action/citation_signature","submit_replication":"https://pith.science/pith/LUS63OLPNFXPQTK4SMY27NKYWC/action/replication_record"}},"created_at":"2026-06-10T01:09:10.838516+00:00","updated_at":"2026-06-10T01:09:10.838516+00:00"}