{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:HSTIKKIZAHS56PVKMJ3I7E64MO","short_pith_number":"pith:HSTIKKIZ","canonical_record":{"source":{"id":"2410.02498","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-03T14:00:44Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"54555bcbef3838efa48753b7c519a549171d97df10fd7d4b316fc5e20b1cc415","abstract_canon_sha256":"f55be18d2f074b14dab30f6d74eb82b02dd3e94af642ba6a8d6ff6f031317964"},"schema_version":"1.0"},"canonical_sha256":"3ca685291901e5df3eaa62768f93dc63bcf02062ac57c963ff0d037cc94a7c6f","source":{"kind":"arxiv","id":"2410.02498","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.02498","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"arxiv_version","alias_value":"2410.02498v1","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.02498","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"pith_short_12","alias_value":"HSTIKKIZAHS5","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"pith_short_16","alias_value":"HSTIKKIZAHS56PVK","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"pith_short_8","alias_value":"HSTIKKIZ","created_at":"2026-07-05T09:15:21Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:HSTIKKIZAHS56PVKMJ3I7E64MO","target":"record","payload":{"canonical_record":{"source":{"id":"2410.02498","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-03T14:00:44Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"54555bcbef3838efa48753b7c519a549171d97df10fd7d4b316fc5e20b1cc415","abstract_canon_sha256":"f55be18d2f074b14dab30f6d74eb82b02dd3e94af642ba6a8d6ff6f031317964"},"schema_version":"1.0"},"canonical_sha256":"3ca685291901e5df3eaa62768f93dc63bcf02062ac57c963ff0d037cc94a7c6f","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:15:21.804142Z","signature_b64":"pBcZkU8+4rCYCjlkz/y5oh493Fq2DlRtDWfohGKxwTD5NIief965ktsimCue161J8DgdWKlXALDCMB94iz0wBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3ca685291901e5df3eaa62768f93dc63bcf02062ac57c963ff0d037cc94a7c6f","last_reissued_at":"2026-07-05T09:15:21.803675Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:15:21.803675Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.02498","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:15:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"4RaIwMj0op1nqEuXFDacuFfzsj/N03zyozmch5P+WbNd5QMszhL6XjC45OrYqcpU7z3LzTacj2nkV17xnfCCAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T01:40:07.907410Z"},"content_sha256":"380b584f6dc421abd8aec169a65848497e104091f2ab6c2b2aa439186279940a","schema_version":"1.0","event_id":"sha256:380b584f6dc421abd8aec169a65848497e104091f2ab6c2b2aa439186279940a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:HSTIKKIZAHS56PVKMJ3I7E64MO","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Dynamic Gradient Alignment for Online Data Mixing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"David Grangier, Pierre Ablin, Simin Fan","submitted_at":"2024-10-03T14:00:44Z","abstract_excerpt":"The composition of training data mixtures is critical for effectively training large language models (LLMs), as it directly impacts their performance on downstream tasks. Our goal is to identify an optimal data mixture to specialize an LLM for a specific task with access to only a few examples. Traditional approaches to this problem include ad-hoc reweighting methods, importance sampling, and gradient alignment techniques. This paper focuses on gradient alignment and introduces Dynamic Gradient Alignment (DGA), a scalable online gradient alignment algorithm. DGA dynamically estimates the pre-t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.02498","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/2410.02498/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:15:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LrP1cs71WaUNXbHJeT9Q4zUeVWA/IRT6l6o3qCiak7xj0XzpxgCVvnTnIQg8Fh0rMl7qleyFORmWrDXg86HqDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T01:40:07.907901Z"},"content_sha256":"7d7b97ef5f3dc2b52184d40c9eba08dced5b206e0d363934607fe1a1fdbe5c75","schema_version":"1.0","event_id":"sha256:7d7b97ef5f3dc2b52184d40c9eba08dced5b206e0d363934607fe1a1fdbe5c75"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/HSTIKKIZAHS56PVKMJ3I7E64MO/bundle.json","state_url":"https://pith.science/pith/HSTIKKIZAHS56PVKMJ3I7E64MO/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/HSTIKKIZAHS56PVKMJ3I7E64MO/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-09T01:40:07Z","links":{"resolver":"https://pith.science/pith/HSTIKKIZAHS56PVKMJ3I7E64MO","bundle":"https://pith.science/pith/HSTIKKIZAHS56PVKMJ3I7E64MO/bundle.json","state":"https://pith.science/pith/HSTIKKIZAHS56PVKMJ3I7E64MO/state.json","well_known_bundle":"https://pith.science/.well-known/pith/HSTIKKIZAHS56PVKMJ3I7E64MO/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:HSTIKKIZAHS56PVKMJ3I7E64MO","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f55be18d2f074b14dab30f6d74eb82b02dd3e94af642ba6a8d6ff6f031317964","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-03T14:00:44Z","title_canon_sha256":"54555bcbef3838efa48753b7c519a549171d97df10fd7d4b316fc5e20b1cc415"},"schema_version":"1.0","source":{"id":"2410.02498","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.02498","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"arxiv_version","alias_value":"2410.02498v1","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.02498","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"pith_short_12","alias_value":"HSTIKKIZAHS5","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"pith_short_16","alias_value":"HSTIKKIZAHS56PVK","created_at":"2026-07-05T09:15:21Z"},{"alias_kind":"pith_short_8","alias_value":"HSTIKKIZ","created_at":"2026-07-05T09:15:21Z"}],"graph_snapshots":[{"event_id":"sha256:7d7b97ef5f3dc2b52184d40c9eba08dced5b206e0d363934607fe1a1fdbe5c75","target":"graph","created_at":"2026-07-05T09:15:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2410.02498/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The composition of training data mixtures is critical for effectively training large language models (LLMs), as it directly impacts their performance on downstream tasks. Our goal is to identify an optimal data mixture to specialize an LLM for a specific task with access to only a few examples. Traditional approaches to this problem include ad-hoc reweighting methods, importance sampling, and gradient alignment techniques. This paper focuses on gradient alignment and introduces Dynamic Gradient Alignment (DGA), a scalable online gradient alignment algorithm. DGA dynamically estimates the pre-t","authors_text":"David Grangier, Pierre Ablin, Simin Fan","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-03T14:00:44Z","title":"Dynamic Gradient Alignment for Online Data Mixing"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.02498","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:380b584f6dc421abd8aec169a65848497e104091f2ab6c2b2aa439186279940a","target":"record","created_at":"2026-07-05T09:15:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f55be18d2f074b14dab30f6d74eb82b02dd3e94af642ba6a8d6ff6f031317964","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-03T14:00:44Z","title_canon_sha256":"54555bcbef3838efa48753b7c519a549171d97df10fd7d4b316fc5e20b1cc415"},"schema_version":"1.0","source":{"id":"2410.02498","kind":"arxiv","version":1}},"canonical_sha256":"3ca685291901e5df3eaa62768f93dc63bcf02062ac57c963ff0d037cc94a7c6f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3ca685291901e5df3eaa62768f93dc63bcf02062ac57c963ff0d037cc94a7c6f","first_computed_at":"2026-07-05T09:15:21.803675Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:15:21.803675Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"pBcZkU8+4rCYCjlkz/y5oh493Fq2DlRtDWfohGKxwTD5NIief965ktsimCue161J8DgdWKlXALDCMB94iz0wBA==","signature_status":"signed_v1","signed_at":"2026-07-05T09:15:21.804142Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.02498","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:380b584f6dc421abd8aec169a65848497e104091f2ab6c2b2aa439186279940a","sha256:7d7b97ef5f3dc2b52184d40c9eba08dced5b206e0d363934607fe1a1fdbe5c75"],"state_sha256":"ef2f8d46d6aaefef61dfeb9d886e4807030016f61ffb4d744f56154af5d36563"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LdUyPlaZFNA7LJpz8252tVx9KBIORSuqUIlxeAptjtG2boOKluyjxRXGftMNBboz32ZTmzDb2Z7wlqlCAf8rAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T01:40:07.912901Z","bundle_sha256":"b18b147f625dcc43ce44262802bd873837c6432b55d55901319f9bb94246158a"}}