{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:WQ3MBGKQNBC3HINPMHCPFO7HXP","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":"26ecc0725eb1d5d23b9592e66c2c085489973cc5e00cd72c346a18720ea05657","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-19T22:03:29Z","title_canon_sha256":"bf9a95b979750584e07a21384d86edfc4ed095fff75093558bfb7eecabe4580f"},"schema_version":"1.0","source":{"id":"2604.17633","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.17633","created_at":"2026-06-29T01:14:31Z"},{"alias_kind":"arxiv_version","alias_value":"2604.17633v2","created_at":"2026-06-29T01:14:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.17633","created_at":"2026-06-29T01:14:31Z"},{"alias_kind":"pith_short_12","alias_value":"WQ3MBGKQNBC3","created_at":"2026-06-29T01:14:31Z"},{"alias_kind":"pith_short_16","alias_value":"WQ3MBGKQNBC3HINP","created_at":"2026-06-29T01:14:31Z"},{"alias_kind":"pith_short_8","alias_value":"WQ3MBGKQ","created_at":"2026-06-29T01:14:31Z"}],"graph_snapshots":[{"event_id":"sha256:d1406fd2a2202b2ee358296b47ed0b4c86539ca6b6fd3274c164ae8e7671b0ca","target":"graph","created_at":"2026-06-29T01:14:31Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We find that the model quickly acquires basic linguistic capabilities in parallel with token-level copying, while translation develops in two distinct phases: an initial phase dominated by copying and surface-level similarities, and a second phase in which more generalizing translation mechanisms are developed while copying is refined."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The behavioral analyses, model-component inspections, and parameter-based ablations on the chosen checkpoints and novel dataset accurately isolate translation dynamics from confounding factors such as data overlap or model architecture specifics."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Multilingual pretraining develops translation in two phases: early copying driven by surface similarities, followed by generalizing mechanisms while copying is refined."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"In multilingual pretraining, models first copy tokens before developing general translation abilities."}],"snapshot_sha256":"d124692a45724e99173aa811e49d23a336ca79d373a5e9e95a5db13da97201ff"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.17633/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models exhibit impressive cross-lingual capabilities. However, prior work analyzes this phenomenon through isolated factors and at sparse points during training, limiting our understanding of how cross-lingual generalization emerges--particularly in the early phases of learning. To study the early trajectory of linguistic and translation capabilities, we pretrain a multilingual 1.7B model on nine diverse languages, capturing checkpoints at a much finer granularity. We use word-level translation as a testbed, introducing a novel dataset to trace how translation develops over trai","authors_text":"Barbara Plank, Felicia K\\\"orner, Florian Eichin, Gitta Kutyniok, Maria Matveev, Michael A. Hedderich","cross_cats":[],"headline":"In multilingual pretraining, models first copy tokens before developing general translation abilities.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-19T22:03:29Z","title":"Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.17633","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T05:33:30.283524Z","id":"d6b52f51-0fc2-47f2-a353-e7cb68dd237d","model_set":{"reader":"grok-4.3"},"one_line_summary":"Multilingual pretraining develops translation in two phases: early copying driven by surface similarities, followed by generalizing mechanisms while copying is refined.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"In multilingual pretraining, models first copy tokens before developing general translation abilities.","strongest_claim":"We find that the model quickly acquires basic linguistic capabilities in parallel with token-level copying, while translation develops in two distinct phases: an initial phase dominated by copying and surface-level similarities, and a second phase in which more generalizing translation mechanisms are developed while copying is refined.","weakest_assumption":"The behavioral analyses, model-component inspections, and parameter-based ablations on the chosen checkpoints and novel dataset accurately isolate translation dynamics from confounding factors such as data overlap or model architecture specifics."}},"verdict_id":"d6b52f51-0fc2-47f2-a353-e7cb68dd237d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f1542d7e42fc50bc572b57e26760bf01dd7ee76728c444741cb32e89415a97eb","target":"record","created_at":"2026-06-29T01:14:31Z","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":"26ecc0725eb1d5d23b9592e66c2c085489973cc5e00cd72c346a18720ea05657","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-19T22:03:29Z","title_canon_sha256":"bf9a95b979750584e07a21384d86edfc4ed095fff75093558bfb7eecabe4580f"},"schema_version":"1.0","source":{"id":"2604.17633","kind":"arxiv","version":2}},"canonical_sha256":"b436c099506845b3a1af61c4f2bbe7bbf7dc94eaf82729b430158303d1b2072b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b436c099506845b3a1af61c4f2bbe7bbf7dc94eaf82729b430158303d1b2072b","first_computed_at":"2026-06-29T01:14:31.301956Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-29T01:14:31.301956Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"kdGRIhuIp6kEVKV+uUHFo/622MlJu+VfrmEask4DHUIZWC8Bsns2+wUN68malsb+TKk8bofYg7MYDY+kaB2yAw==","signature_status":"signed_v1","signed_at":"2026-06-29T01:14:31.302372Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.17633","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f1542d7e42fc50bc572b57e26760bf01dd7ee76728c444741cb32e89415a97eb","sha256:d1406fd2a2202b2ee358296b47ed0b4c86539ca6b6fd3274c164ae8e7671b0ca"],"state_sha256":"16e273a5d5edea63aef0177fff61963c062a2b5d200d03b6e41b069c06498a73"}