{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:F2CZHFXK6O2B2I7TIBPU4NGAUI","short_pith_number":"pith:F2CZHFXK","schema_version":"1.0","canonical_sha256":"2e859396eaf3b41d23f3405f4e34c0a227583fda214785942689eb65607b5468","source":{"kind":"arxiv","id":"2411.15661","version":2},"attestation_state":"computed","paper":{"title":"Improving Next Tokens via Second-to-Last Predictions with Generate and Refine","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Johannes Schneider","submitted_at":"2024-11-23T22:09:58Z","abstract_excerpt":"Autoregressive language models like GPT aim to predict next tokens, while autoencoding models such as BERT are trained on tasks such as predicting masked tokens. We train a decoder-only architecture for predicting the second to last token for a sequence of tokens. Our approach yields higher computational training efficiency than BERT-style models by employing a structured deterministic approach to masking tokens. We use our model to improve the next token predictions of a standard GPT by combining both predictions in a ``generate-then-refine'' approach. We demonstrate on different variants of "},"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":"2411.15661","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-23T22:09:58Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b9f7694ed4ec545612a476b606a66df0889f37d53e3163e419b1c52c5c8abe8e","abstract_canon_sha256":"94a512d9769d71c544b9e15d41cc82e1269c0b891389d9308d7b5c918a9293f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:14:07.155800Z","signature_b64":"HOuh+jUT8V7nY8zFdCf+wmD0xVMTip9s1lqj7CbG+PaKc8lzlEnXQ5+t9jwKzHcX0dCVMCJwuTdlkoZbWj1dCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2e859396eaf3b41d23f3405f4e34c0a227583fda214785942689eb65607b5468","last_reissued_at":"2026-07-05T10:14:07.155323Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:14:07.155323Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving Next Tokens via Second-to-Last Predictions with Generate and Refine","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Johannes Schneider","submitted_at":"2024-11-23T22:09:58Z","abstract_excerpt":"Autoregressive language models like GPT aim to predict next tokens, while autoencoding models such as BERT are trained on tasks such as predicting masked tokens. We train a decoder-only architecture for predicting the second to last token for a sequence of tokens. Our approach yields higher computational training efficiency than BERT-style models by employing a structured deterministic approach to masking tokens. We use our model to improve the next token predictions of a standard GPT by combining both predictions in a ``generate-then-refine'' approach. We demonstrate on different variants of "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.15661","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/2411.15661/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":"2411.15661","created_at":"2026-07-05T10:14:07.155385+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.15661v2","created_at":"2026-07-05T10:14:07.155385+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.15661","created_at":"2026-07-05T10:14:07.155385+00:00"},{"alias_kind":"pith_short_12","alias_value":"F2CZHFXK6O2B","created_at":"2026-07-05T10:14:07.155385+00:00"},{"alias_kind":"pith_short_16","alias_value":"F2CZHFXK6O2B2I7T","created_at":"2026-07-05T10:14:07.155385+00:00"},{"alias_kind":"pith_short_8","alias_value":"F2CZHFXK","created_at":"2026-07-05T10:14:07.155385+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/F2CZHFXK6O2B2I7TIBPU4NGAUI","json":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI.json","graph_json":"https://pith.science/api/pith-number/F2CZHFXK6O2B2I7TIBPU4NGAUI/graph.json","events_json":"https://pith.science/api/pith-number/F2CZHFXK6O2B2I7TIBPU4NGAUI/events.json","paper":"https://pith.science/paper/F2CZHFXK"},"agent_actions":{"view_html":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI","download_json":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI.json","view_paper":"https://pith.science/paper/F2CZHFXK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.15661&json=true","fetch_graph":"https://pith.science/api/pith-number/F2CZHFXK6O2B2I7TIBPU4NGAUI/graph.json","fetch_events":"https://pith.science/api/pith-number/F2CZHFXK6O2B2I7TIBPU4NGAUI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI/action/storage_attestation","attest_author":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI/action/author_attestation","sign_citation":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI/action/citation_signature","submit_replication":"https://pith.science/pith/F2CZHFXK6O2B2I7TIBPU4NGAUI/action/replication_record"}},"created_at":"2026-07-05T10:14:07.155385+00:00","updated_at":"2026-07-05T10:14:07.155385+00:00"}