{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:I6PXG2LGJB3372QY5NGPDSEWWU","short_pith_number":"pith:I6PXG2LG","schema_version":"1.0","canonical_sha256":"479f7369664877bfea18eb4cf1c896b53337714116cf731106bd2b76f258c07c","source":{"kind":"arxiv","id":"2504.19333","version":2},"attestation_state":"computed","paper":{"title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Eric Lin, James O' Neill, Santhosh Subramanian, Vaikkunth Mugunthan","submitted_at":"2025-04-27T19:07:58Z","abstract_excerpt":"The trend towards large language models (LLMs) for guardrailing against undesired behaviors is increasing and has shown promise for censoring user inputs. However, increased latency, memory consumption, hosting expenses and non-structured outputs can make their use prohibitive.\n  In this work, we show that task-specific data generation can lead to fine-tuned classifiers that significantly outperform current state of the art (SoTA) while being orders of magnitude smaller. Secondly, we show that using a single model, \\texttt{MultiTaskGuard}, that is pretrained on a large synthetically generated "},"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":"2504.19333","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-27T19:07:58Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"f742b5e27c0cdd41e5eeb92ac74e9df0a2c5b9707d8b2bb17e533d38e07b3279","abstract_canon_sha256":"85682e730990c54349ed69c223c9650d9c475db0f668932044c25af408087f98"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:55:35.681627Z","signature_b64":"geeNFpTidXjDnpdpBanHE7+n+avJ9xuKpVH/Qr46Mf32Xyxv2rM0boDGUcAcL3KXUsnb6VYtgeNLui4FkC63BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"479f7369664877bfea18eb4cf1c896b53337714116cf731106bd2b76f258c07c","last_reissued_at":"2026-07-05T10:55:35.681052Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:55:35.681052Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Eric Lin, James O' Neill, Santhosh Subramanian, Vaikkunth Mugunthan","submitted_at":"2025-04-27T19:07:58Z","abstract_excerpt":"The trend towards large language models (LLMs) for guardrailing against undesired behaviors is increasing and has shown promise for censoring user inputs. However, increased latency, memory consumption, hosting expenses and non-structured outputs can make their use prohibitive.\n  In this work, we show that task-specific data generation can lead to fine-tuned classifiers that significantly outperform current state of the art (SoTA) while being orders of magnitude smaller. Secondly, we show that using a single model, \\texttt{MultiTaskGuard}, that is pretrained on a large synthetically generated "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.19333","kind":"arxiv","version":2},"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/2504.19333/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":"2504.19333","created_at":"2026-07-05T10:55:35.681107+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.19333v2","created_at":"2026-07-05T10:55:35.681107+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.19333","created_at":"2026-07-05T10:55:35.681107+00:00"},{"alias_kind":"pith_short_12","alias_value":"I6PXG2LGJB33","created_at":"2026-07-05T10:55:35.681107+00:00"},{"alias_kind":"pith_short_16","alias_value":"I6PXG2LGJB3372QY","created_at":"2026-07-05T10:55:35.681107+00:00"},{"alias_kind":"pith_short_8","alias_value":"I6PXG2LG","created_at":"2026-07-05T10:55:35.681107+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/I6PXG2LGJB3372QY5NGPDSEWWU","json":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU.json","graph_json":"https://pith.science/api/pith-number/I6PXG2LGJB3372QY5NGPDSEWWU/graph.json","events_json":"https://pith.science/api/pith-number/I6PXG2LGJB3372QY5NGPDSEWWU/events.json","paper":"https://pith.science/paper/I6PXG2LG"},"agent_actions":{"view_html":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU","download_json":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU.json","view_paper":"https://pith.science/paper/I6PXG2LG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.19333&json=true","fetch_graph":"https://pith.science/api/pith-number/I6PXG2LGJB3372QY5NGPDSEWWU/graph.json","fetch_events":"https://pith.science/api/pith-number/I6PXG2LGJB3372QY5NGPDSEWWU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU/action/storage_attestation","attest_author":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU/action/author_attestation","sign_citation":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU/action/citation_signature","submit_replication":"https://pith.science/pith/I6PXG2LGJB3372QY5NGPDSEWWU/action/replication_record"}},"created_at":"2026-07-05T10:55:35.681107+00:00","updated_at":"2026-07-05T10:55:35.681107+00:00"}