{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:STJSUF57RL7CNTVXGDHDL6NDL3","short_pith_number":"pith:STJSUF57","schema_version":"1.0","canonical_sha256":"94d32a17bf8afe26ceb730ce35f9a35ed62dbd0725057e4b6e2e5a43fbe109a1","source":{"kind":"arxiv","id":"2505.10518","version":1},"attestation_state":"computed","paper":{"title":"Multi-Token Prediction Needs Registers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anastasios Gerontopoulos, Nikos Komodakis, Spyros Gidaris","submitted_at":"2025-05-15T17:25:03Z","abstract_excerpt":"Multi-token prediction has emerged as a promising objective for improving language model pretraining, but its benefits have not consistently generalized to other settings such as fine-tuning. In this paper, we propose MuToR, a simple and effective approach to multi-token prediction that interleaves learnable register tokens into the input sequence, each tasked with predicting future targets. Compared to existing methods, MuToR offers several key advantages: it introduces only a negligible number of additional parameters, requires no architectural changes--ensuring compatibility with off-the-sh"},"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":"2505.10518","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-15T17:25:03Z","cross_cats_sorted":["cs.AI","cs.CV","cs.LG"],"title_canon_sha256":"f5a499ae40f41ff1b11c5f951b31468369bdbbd9bb7f68136eb7c20e249b6662","abstract_canon_sha256":"18047b6b356a5b38f87c5554edc93e716af921b6ac1604d2975ab385a3a5901a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:03:39.362728Z","signature_b64":"mX1djC4FuaxbxatpGrOk8sMAAiDZSw5M+weZhzdgMS0h2XLhBLOKYj0/LgW/kzfPpMZHAWkhL36KjnceafuhDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"94d32a17bf8afe26ceb730ce35f9a35ed62dbd0725057e4b6e2e5a43fbe109a1","last_reissued_at":"2026-07-05T11:03:39.362230Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:03:39.362230Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Token Prediction Needs Registers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anastasios Gerontopoulos, Nikos Komodakis, Spyros Gidaris","submitted_at":"2025-05-15T17:25:03Z","abstract_excerpt":"Multi-token prediction has emerged as a promising objective for improving language model pretraining, but its benefits have not consistently generalized to other settings such as fine-tuning. In this paper, we propose MuToR, a simple and effective approach to multi-token prediction that interleaves learnable register tokens into the input sequence, each tasked with predicting future targets. Compared to existing methods, MuToR offers several key advantages: it introduces only a negligible number of additional parameters, requires no architectural changes--ensuring compatibility with off-the-sh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.10518","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/2505.10518/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":"2505.10518","created_at":"2026-07-05T11:03:39.362289+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.10518v1","created_at":"2026-07-05T11:03:39.362289+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.10518","created_at":"2026-07-05T11:03:39.362289+00:00"},{"alias_kind":"pith_short_12","alias_value":"STJSUF57RL7C","created_at":"2026-07-05T11:03:39.362289+00:00"},{"alias_kind":"pith_short_16","alias_value":"STJSUF57RL7CNTVX","created_at":"2026-07-05T11:03:39.362289+00:00"},{"alias_kind":"pith_short_8","alias_value":"STJSUF57","created_at":"2026-07-05T11:03:39.362289+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2509.01649","citing_title":"Distilled Pretraining: A modern lens of Data, In-Context Learning and Test-Time Scaling","ref_index":29,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3","json":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3.json","graph_json":"https://pith.science/api/pith-number/STJSUF57RL7CNTVXGDHDL6NDL3/graph.json","events_json":"https://pith.science/api/pith-number/STJSUF57RL7CNTVXGDHDL6NDL3/events.json","paper":"https://pith.science/paper/STJSUF57"},"agent_actions":{"view_html":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3","download_json":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3.json","view_paper":"https://pith.science/paper/STJSUF57","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.10518&json=true","fetch_graph":"https://pith.science/api/pith-number/STJSUF57RL7CNTVXGDHDL6NDL3/graph.json","fetch_events":"https://pith.science/api/pith-number/STJSUF57RL7CNTVXGDHDL6NDL3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3/action/storage_attestation","attest_author":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3/action/author_attestation","sign_citation":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3/action/citation_signature","submit_replication":"https://pith.science/pith/STJSUF57RL7CNTVXGDHDL6NDL3/action/replication_record"}},"created_at":"2026-07-05T11:03:39.362289+00:00","updated_at":"2026-07-05T11:03:39.362289+00:00"}