{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FSBVO47YPF2NT3N5E4UNCMX2HD","short_pith_number":"pith:FSBVO47Y","schema_version":"1.0","canonical_sha256":"2c835773f87974d9edbd2728d132fa38db327fab110c3dcf17b961126d487390","source":{"kind":"arxiv","id":"2502.12120","version":3},"attestation_state":"computed","paper":{"title":"LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Matthias Bethge, Prasanna Mayilvahanan, Sayak Mallick, Thadd\\\"aus Wiedemer, Wieland Brendel","submitted_at":"2025-02-17T18:45:25Z","abstract_excerpt":"Scaling laws guide the development of large language models (LLMs) by offering estimates for the optimal balance of model size, tokens, and compute. More recently, loss-to-loss scaling laws that relate losses across pretraining datasets and downstream tasks have emerged as a powerful tool for understanding and improving LLM performance and generalization. In this work, we investigate which factors most strongly influence loss-to-loss scaling. Our experiments reveal that the pretraining data determines the scaling trend. In contrast, model size, optimization hyperparameters, tokenizer and even "},"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":"2502.12120","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-17T18:45:25Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"cc4618a395ac64e37ac649fa3515b9697d3aebedf057c2f666392e8ec6f108e8","abstract_canon_sha256":"0d10e906da89724f7de4cf358c471a9eb074ae8adb67ebe3820934bbcb338e12"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:05:04.409365Z","signature_b64":"0Vs+Z1wQamSxQXiPKdSU1FZRUQmeNd56cBnSHBLsCpTijXeuAsTv6/e/wj7sNVhh5seYrzOuayEL5UY+eRLhDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2c835773f87974d9edbd2728d132fa38db327fab110c3dcf17b961126d487390","last_reissued_at":"2026-05-21T01:05:04.408522Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:05:04.408522Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Matthias Bethge, Prasanna Mayilvahanan, Sayak Mallick, Thadd\\\"aus Wiedemer, Wieland Brendel","submitted_at":"2025-02-17T18:45:25Z","abstract_excerpt":"Scaling laws guide the development of large language models (LLMs) by offering estimates for the optimal balance of model size, tokens, and compute. More recently, loss-to-loss scaling laws that relate losses across pretraining datasets and downstream tasks have emerged as a powerful tool for understanding and improving LLM performance and generalization. In this work, we investigate which factors most strongly influence loss-to-loss scaling. Our experiments reveal that the pretraining data determines the scaling trend. In contrast, model size, optimization hyperparameters, tokenizer and even "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12120","kind":"arxiv","version":3},"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/2502.12120/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":"2502.12120","created_at":"2026-05-21T01:05:04.408663+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.12120v3","created_at":"2026-05-21T01:05:04.408663+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12120","created_at":"2026-05-21T01:05:04.408663+00:00"},{"alias_kind":"pith_short_12","alias_value":"FSBVO47YPF2N","created_at":"2026-05-21T01:05:04.408663+00:00"},{"alias_kind":"pith_short_16","alias_value":"FSBVO47YPF2NT3N5","created_at":"2026-05-21T01:05:04.408663+00:00"},{"alias_kind":"pith_short_8","alias_value":"FSBVO47Y","created_at":"2026-05-21T01:05:04.408663+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/FSBVO47YPF2NT3N5E4UNCMX2HD","json":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD.json","graph_json":"https://pith.science/api/pith-number/FSBVO47YPF2NT3N5E4UNCMX2HD/graph.json","events_json":"https://pith.science/api/pith-number/FSBVO47YPF2NT3N5E4UNCMX2HD/events.json","paper":"https://pith.science/paper/FSBVO47Y"},"agent_actions":{"view_html":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD","download_json":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD.json","view_paper":"https://pith.science/paper/FSBVO47Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.12120&json=true","fetch_graph":"https://pith.science/api/pith-number/FSBVO47YPF2NT3N5E4UNCMX2HD/graph.json","fetch_events":"https://pith.science/api/pith-number/FSBVO47YPF2NT3N5E4UNCMX2HD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD/action/storage_attestation","attest_author":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD/action/author_attestation","sign_citation":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD/action/citation_signature","submit_replication":"https://pith.science/pith/FSBVO47YPF2NT3N5E4UNCMX2HD/action/replication_record"}},"created_at":"2026-05-21T01:05:04.408663+00:00","updated_at":"2026-05-21T01:05:04.408663+00:00"}