{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:752I57P3QIYKNKLB5P5N5KDJ3H","short_pith_number":"pith:752I57P3","schema_version":"1.0","canonical_sha256":"ff748efdfb8230a6a961ebfadea869d9f125888e7feb7e31e00fde78d217e735","source":{"kind":"arxiv","id":"2012.11635","version":2},"attestation_state":"computed","paper":{"title":"A Distributional Approach to Controlled Text Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hady Elsahar, Marc Dymetman, Muhammad Khalifa","submitted_at":"2020-12-21T19:02:41Z","abstract_excerpt":"We propose a Distributional Approach for addressing Controlled Text Generation from pre-trained Language Models (LMs). This approach permits to specify, in a single formal framework, both \"pointwise\" and \"distributional\" constraints over the target LM -- to our knowledge, the first model with such generality -- while minimizing KL divergence from the initial LM distribution. The optimal target distribution is then uniquely determined as an explicit EBM (Energy-Based Model) representation. From that optimal representation we then train a target controlled Autoregressive LM through an adaptive d"},"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":"2012.11635","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2020-12-21T19:02:41Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"069f3f9b3da554e24b82381b56cb5b2d6cfbb0d50897d698fef36034f064abcc","abstract_canon_sha256":"a2cc9aeb9804163de03cfa437973eb25f9e1d5e5a97fe444b4c9eb232c18a5b5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:37:57.433802Z","signature_b64":"hTOv/z2HcLDOaitA1PKJ2l8/TL0QEvchqgWIWprYy7wpzeHaz6KnRnheM4IEJmr05oD+oUPChgOnzXu/VQExAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ff748efdfb8230a6a961ebfadea869d9f125888e7feb7e31e00fde78d217e735","last_reissued_at":"2026-07-05T02:37:57.433359Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:37:57.433359Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Distributional Approach to Controlled Text Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hady Elsahar, Marc Dymetman, Muhammad Khalifa","submitted_at":"2020-12-21T19:02:41Z","abstract_excerpt":"We propose a Distributional Approach for addressing Controlled Text Generation from pre-trained Language Models (LMs). This approach permits to specify, in a single formal framework, both \"pointwise\" and \"distributional\" constraints over the target LM -- to our knowledge, the first model with such generality -- while minimizing KL divergence from the initial LM distribution. The optimal target distribution is then uniquely determined as an explicit EBM (Energy-Based Model) representation. From that optimal representation we then train a target controlled Autoregressive LM through an adaptive d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2012.11635","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/2012.11635/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":"2012.11635","created_at":"2026-07-05T02:37:57.433422+00:00"},{"alias_kind":"arxiv_version","alias_value":"2012.11635v2","created_at":"2026-07-05T02:37:57.433422+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2012.11635","created_at":"2026-07-05T02:37:57.433422+00:00"},{"alias_kind":"pith_short_12","alias_value":"752I57P3QIYK","created_at":"2026-07-05T02:37:57.433422+00:00"},{"alias_kind":"pith_short_16","alias_value":"752I57P3QIYKNKLB","created_at":"2026-07-05T02:37:57.433422+00:00"},{"alias_kind":"pith_short_8","alias_value":"752I57P3","created_at":"2026-07-05T02:37:57.433422+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":114,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12288","citing_title":"TokenRatio: Principled Token-Level Preference Optimization via Ratio Matching","ref_index":114,"is_internal_anchor":false},{"citing_arxiv_id":"2112.04359","citing_title":"Ethical and social risks of harm from Language Models","ref_index":142,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H","json":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H.json","graph_json":"https://pith.science/api/pith-number/752I57P3QIYKNKLB5P5N5KDJ3H/graph.json","events_json":"https://pith.science/api/pith-number/752I57P3QIYKNKLB5P5N5KDJ3H/events.json","paper":"https://pith.science/paper/752I57P3"},"agent_actions":{"view_html":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H","download_json":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H.json","view_paper":"https://pith.science/paper/752I57P3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2012.11635&json=true","fetch_graph":"https://pith.science/api/pith-number/752I57P3QIYKNKLB5P5N5KDJ3H/graph.json","fetch_events":"https://pith.science/api/pith-number/752I57P3QIYKNKLB5P5N5KDJ3H/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H/action/timestamp_anchor","attest_storage":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H/action/storage_attestation","attest_author":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H/action/author_attestation","sign_citation":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H/action/citation_signature","submit_replication":"https://pith.science/pith/752I57P3QIYKNKLB5P5N5KDJ3H/action/replication_record"}},"created_at":"2026-07-05T02:37:57.433422+00:00","updated_at":"2026-07-05T02:37:57.433422+00:00"}