{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:VALTRU2QBHULPUXEOM4HTJECET","short_pith_number":"pith:VALTRU2Q","schema_version":"1.0","canonical_sha256":"a81738d35009e8b7d2e4733879a48224e5ff6fb141f593e9d52c605c4ffa0db0","source":{"kind":"arxiv","id":"2403.04510","version":1},"attestation_state":"computed","paper":{"title":"Where does In-context Translation Happen in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Mueller, Kevin Duh, Suzanna Sia","submitted_at":"2024-03-07T14:12:41Z","abstract_excerpt":"Self-supervised large language models have demonstrated the ability to perform Machine Translation (MT) via in-context learning, but little is known about where the model performs the task with respect to prompt instructions and demonstration examples. In this work, we attempt to characterize the region where large language models transition from in-context learners to translation models. Through a series of layer-wise context-masking experiments on \\textsc{GPTNeo2.7B}, \\textsc{Bloom3B}, \\textsc{Llama7b} and \\textsc{Llama7b-chat}, we demonstrate evidence of a \"task recognition\" point where the"},"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":"2403.04510","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-07T14:12:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"61a4250f35e27a158bd030d31129f361c0e58932afd3b561e0d3ba3d8d28fd66","abstract_canon_sha256":"061b8dd75a26b6673b04bd454584295b39fbfa3db0fd06cfeb55fe1f14b85128"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:53:24.367947Z","signature_b64":"KcgUigOc2tfJEyNYxI1B5PC1w5nlUunF5MFS3OsCsjSvo7o/7k2XQ+VSwIUyXRTHbMU+HBnxutoDGnoYWt5eDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a81738d35009e8b7d2e4733879a48224e5ff6fb141f593e9d52c605c4ffa0db0","last_reissued_at":"2026-07-05T07:53:24.367374Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:53:24.367374Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Where does In-context Translation Happen in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Mueller, Kevin Duh, Suzanna Sia","submitted_at":"2024-03-07T14:12:41Z","abstract_excerpt":"Self-supervised large language models have demonstrated the ability to perform Machine Translation (MT) via in-context learning, but little is known about where the model performs the task with respect to prompt instructions and demonstration examples. In this work, we attempt to characterize the region where large language models transition from in-context learners to translation models. Through a series of layer-wise context-masking experiments on \\textsc{GPTNeo2.7B}, \\textsc{Bloom3B}, \\textsc{Llama7b} and \\textsc{Llama7b-chat}, we demonstrate evidence of a \"task recognition\" point where the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.04510","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/2403.04510/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":"2403.04510","created_at":"2026-07-05T07:53:24.367437+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.04510v1","created_at":"2026-07-05T07:53:24.367437+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.04510","created_at":"2026-07-05T07:53:24.367437+00:00"},{"alias_kind":"pith_short_12","alias_value":"VALTRU2QBHUL","created_at":"2026-07-05T07:53:24.367437+00:00"},{"alias_kind":"pith_short_16","alias_value":"VALTRU2QBHULPUXE","created_at":"2026-07-05T07:53:24.367437+00:00"},{"alias_kind":"pith_short_8","alias_value":"VALTRU2Q","created_at":"2026-07-05T07:53:24.367437+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06906","citing_title":"EASE-TTT: Evidence-Aligned Selective Test-Time Training for Long-Context Question Answering","ref_index":68,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET","json":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET.json","graph_json":"https://pith.science/api/pith-number/VALTRU2QBHULPUXEOM4HTJECET/graph.json","events_json":"https://pith.science/api/pith-number/VALTRU2QBHULPUXEOM4HTJECET/events.json","paper":"https://pith.science/paper/VALTRU2Q"},"agent_actions":{"view_html":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET","download_json":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET.json","view_paper":"https://pith.science/paper/VALTRU2Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.04510&json=true","fetch_graph":"https://pith.science/api/pith-number/VALTRU2QBHULPUXEOM4HTJECET/graph.json","fetch_events":"https://pith.science/api/pith-number/VALTRU2QBHULPUXEOM4HTJECET/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET/action/storage_attestation","attest_author":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET/action/author_attestation","sign_citation":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET/action/citation_signature","submit_replication":"https://pith.science/pith/VALTRU2QBHULPUXEOM4HTJECET/action/replication_record"}},"created_at":"2026-07-05T07:53:24.367437+00:00","updated_at":"2026-07-05T07:53:24.367437+00:00"}