{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:47VQ2MJTMQD7V55YOZQECZRWL3","short_pith_number":"pith:47VQ2MJT","schema_version":"1.0","canonical_sha256":"e7eb0d31336407faf7b876604166365ed5e735e93661b3377c2973bfda891922","source":{"kind":"arxiv","id":"2110.08246","version":1},"attestation_state":"computed","paper":{"title":"Tricks for Training Sparse Translation Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Angela Fan, Dheeru Dua, James Cross, Mike Lewis, Shruti Bhosale, Vedanuj Goswami","submitted_at":"2021-10-15T17:58:45Z","abstract_excerpt":"Multi-task learning with an unbalanced data distribution skews model learning towards high resource tasks, especially when model capacity is fixed and fully shared across all tasks. Sparse scaling architectures, such as BASELayers, provide flexible mechanisms for different tasks to have a variable number of parameters, which can be useful to counterbalance skewed data distributions. We find that that sparse architectures for multilingual machine translation can perform poorly out of the box, and propose two straightforward techniques to mitigate this - a temperature heating mechanism and dense"},"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":"2110.08246","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2021-10-15T17:58:45Z","cross_cats_sorted":[],"title_canon_sha256":"b7102a49999bd495e9dc5510671934d7de37b5addf991c2319bcccadce544495","abstract_canon_sha256":"d17a70d7dd63b720310940d06af9b8807a578edb86aa7cc69d75bcccff702d7c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:23:01.836469Z","signature_b64":"zetmTeQBfs1tU5leWE+vVLX3+if6pf3oCGOeM0qZFw4gy/X4TkF70sAxFdH8WzrGd4HVbIeZydgBdarjDvb8DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e7eb0d31336407faf7b876604166365ed5e735e93661b3377c2973bfda891922","last_reissued_at":"2026-07-05T03:23:01.835987Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:23:01.835987Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tricks for Training Sparse Translation Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Angela Fan, Dheeru Dua, James Cross, Mike Lewis, Shruti Bhosale, Vedanuj Goswami","submitted_at":"2021-10-15T17:58:45Z","abstract_excerpt":"Multi-task learning with an unbalanced data distribution skews model learning towards high resource tasks, especially when model capacity is fixed and fully shared across all tasks. Sparse scaling architectures, such as BASELayers, provide flexible mechanisms for different tasks to have a variable number of parameters, which can be useful to counterbalance skewed data distributions. We find that that sparse architectures for multilingual machine translation can perform poorly out of the box, and propose two straightforward techniques to mitigate this - a temperature heating mechanism and dense"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.08246","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/2110.08246/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":"2110.08246","created_at":"2026-07-05T03:23:01.836044+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.08246v1","created_at":"2026-07-05T03:23:01.836044+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.08246","created_at":"2026-07-05T03:23:01.836044+00:00"},{"alias_kind":"pith_short_12","alias_value":"47VQ2MJTMQD7","created_at":"2026-07-05T03:23:01.836044+00:00"},{"alias_kind":"pith_short_16","alias_value":"47VQ2MJTMQD7V55Y","created_at":"2026-07-05T03:23:01.836044+00:00"},{"alias_kind":"pith_short_8","alias_value":"47VQ2MJT","created_at":"2026-07-05T03:23:01.836044+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.05564","citing_title":"TabICL: A Tabular Foundation Model for In-Context Learning on Large Data","ref_index":127,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3","json":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3.json","graph_json":"https://pith.science/api/pith-number/47VQ2MJTMQD7V55YOZQECZRWL3/graph.json","events_json":"https://pith.science/api/pith-number/47VQ2MJTMQD7V55YOZQECZRWL3/events.json","paper":"https://pith.science/paper/47VQ2MJT"},"agent_actions":{"view_html":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3","download_json":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3.json","view_paper":"https://pith.science/paper/47VQ2MJT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.08246&json=true","fetch_graph":"https://pith.science/api/pith-number/47VQ2MJTMQD7V55YOZQECZRWL3/graph.json","fetch_events":"https://pith.science/api/pith-number/47VQ2MJTMQD7V55YOZQECZRWL3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3/action/storage_attestation","attest_author":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3/action/author_attestation","sign_citation":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3/action/citation_signature","submit_replication":"https://pith.science/pith/47VQ2MJTMQD7V55YOZQECZRWL3/action/replication_record"}},"created_at":"2026-07-05T03:23:01.836044+00:00","updated_at":"2026-07-05T03:23:01.836044+00:00"}