{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:XB7NESBVQTPXQMQ4FMI2K2O467","short_pith_number":"pith:XB7NESBV","schema_version":"1.0","canonical_sha256":"b87ed2483584df78321c2b11a569dcf7d6ce528fc72f1ca53a7abc310e62cc4f","source":{"kind":"arxiv","id":"2004.02644","version":3},"attestation_state":"computed","paper":{"title":"Sparse Text Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andr\\'e F. T. Martins, Pedro Henrique Martins, Zita Marinho","submitted_at":"2020-04-06T13:09:10Z","abstract_excerpt":"Current state-of-the-art text generators build on powerful language models such as GPT-2, achieving impressive performance. However, to avoid degenerate text, they require sampling from a modified softmax, via temperature parameters or ad-hoc truncation techniques, as in top-$k$ or nucleus sampling. This creates a mismatch between training and testing conditions. In this paper, we use the recently introduced entmax transformation to train and sample from a natively sparse language model, avoiding this mismatch. The result is a text generator with favorable performance in terms of fluency and c"},"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":"2004.02644","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-06T13:09:10Z","cross_cats_sorted":[],"title_canon_sha256":"f998b830acbaaba03940a56b2565464cb22c84b64822a62f7fc5752f8cb35033","abstract_canon_sha256":"4bb86fd41effdcad5ddb3d4915bf3eb1610370da9f0fd522a155a62df0e0d7d2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:39:57.212694Z","signature_b64":"WLp2A/p84vjgfbFtvKXFl3lRpK8YbAnWmb7EXooOBnz5ppM+lk7qgtuj2v/lpHBmkaHtymNcF2UH4vu/4aNUCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b87ed2483584df78321c2b11a569dcf7d6ce528fc72f1ca53a7abc310e62cc4f","last_reissued_at":"2026-07-05T01:39:57.212339Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:39:57.212339Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sparse Text Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andr\\'e F. T. Martins, Pedro Henrique Martins, Zita Marinho","submitted_at":"2020-04-06T13:09:10Z","abstract_excerpt":"Current state-of-the-art text generators build on powerful language models such as GPT-2, achieving impressive performance. However, to avoid degenerate text, they require sampling from a modified softmax, via temperature parameters or ad-hoc truncation techniques, as in top-$k$ or nucleus sampling. This creates a mismatch between training and testing conditions. In this paper, we use the recently introduced entmax transformation to train and sample from a natively sparse language model, avoiding this mismatch. The result is a text generator with favorable performance in terms of fluency and c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.02644","kind":"arxiv","version":3},"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/2004.02644/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":"2004.02644","created_at":"2026-07-05T01:39:57.212401+00:00"},{"alias_kind":"arxiv_version","alias_value":"2004.02644v3","created_at":"2026-07-05T01:39:57.212401+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.02644","created_at":"2026-07-05T01:39:57.212401+00:00"},{"alias_kind":"pith_short_12","alias_value":"XB7NESBVQTPX","created_at":"2026-07-05T01:39:57.212401+00:00"},{"alias_kind":"pith_short_16","alias_value":"XB7NESBVQTPXQMQ4","created_at":"2026-07-05T01:39:57.212401+00:00"},{"alias_kind":"pith_short_8","alias_value":"XB7NESBV","created_at":"2026-07-05T01:39:57.212401+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07135","citing_title":"Sparse Attention for Dense Open-Vocabulary Prediction in CLIP","ref_index":16,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467","json":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467.json","graph_json":"https://pith.science/api/pith-number/XB7NESBVQTPXQMQ4FMI2K2O467/graph.json","events_json":"https://pith.science/api/pith-number/XB7NESBVQTPXQMQ4FMI2K2O467/events.json","paper":"https://pith.science/paper/XB7NESBV"},"agent_actions":{"view_html":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467","download_json":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467.json","view_paper":"https://pith.science/paper/XB7NESBV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2004.02644&json=true","fetch_graph":"https://pith.science/api/pith-number/XB7NESBVQTPXQMQ4FMI2K2O467/graph.json","fetch_events":"https://pith.science/api/pith-number/XB7NESBVQTPXQMQ4FMI2K2O467/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467/action/storage_attestation","attest_author":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467/action/author_attestation","sign_citation":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467/action/citation_signature","submit_replication":"https://pith.science/pith/XB7NESBVQTPXQMQ4FMI2K2O467/action/replication_record"}},"created_at":"2026-07-05T01:39:57.212401+00:00","updated_at":"2026-07-05T01:39:57.212401+00:00"}