{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:4P5P4J65QSIF7N3MCWJFFFDWTG","short_pith_number":"pith:4P5P4J65","schema_version":"1.0","canonical_sha256":"e3fafe27dd84905fb76c15925294769996c80babdab92871cbcaace6c96cd70d","source":{"kind":"arxiv","id":"2106.02736","version":2},"attestation_state":"computed","paper":{"title":"Exposing the Implicit Energy Networks behind Masked Language Models via Metropolis--Hastings","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chris Dyer, Kartik Goyal, Taylor Berg-Kirkpatrick","submitted_at":"2021-06-04T22:04:30Z","abstract_excerpt":"While recent work has shown that scores from models trained by the ubiquitous masked language modeling (MLM) objective effectively discriminate probable from improbable sequences, it is still an open question if these MLMs specify a principled probability distribution over the space of possible sequences. In this paper, we interpret MLMs as energy-based sequence models and propose two energy parametrizations derivable from the trained MLMs. In order to draw samples correctly from these models, we develop a tractable sampling scheme based on the Metropolis--Hastings Monte Carlo algorithm. In ou"},"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":"2106.02736","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-04T22:04:30Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"1f5c916f2257e2183e86e6bfa6539e1484229919da690385970221d22b492202","abstract_canon_sha256":"b6d4eead8eb79e96c2be7b5668442417a9be6c30335646c79f510fb767c1eb60"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:05:11.610523Z","signature_b64":"eODY7mHjAPg5tMoP8XWENiQKIf1v48XUMxSRK3wHm5u7niQkTL1lNF2CS8wjPdcTBlzs7fPoPP7Grgz1GOgzBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e3fafe27dd84905fb76c15925294769996c80babdab92871cbcaace6c96cd70d","last_reissued_at":"2026-07-05T04:05:11.610021Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:05:11.610021Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Exposing the Implicit Energy Networks behind Masked Language Models via Metropolis--Hastings","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chris Dyer, Kartik Goyal, Taylor Berg-Kirkpatrick","submitted_at":"2021-06-04T22:04:30Z","abstract_excerpt":"While recent work has shown that scores from models trained by the ubiquitous masked language modeling (MLM) objective effectively discriminate probable from improbable sequences, it is still an open question if these MLMs specify a principled probability distribution over the space of possible sequences. In this paper, we interpret MLMs as energy-based sequence models and propose two energy parametrizations derivable from the trained MLMs. In order to draw samples correctly from these models, we develop a tractable sampling scheme based on the Metropolis--Hastings Monte Carlo algorithm. In ou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.02736","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/2106.02736/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":"2106.02736","created_at":"2026-07-05T04:05:11.610082+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.02736v2","created_at":"2026-07-05T04:05:11.610082+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.02736","created_at":"2026-07-05T04:05:11.610082+00:00"},{"alias_kind":"pith_short_12","alias_value":"4P5P4J65QSIF","created_at":"2026-07-05T04:05:11.610082+00:00"},{"alias_kind":"pith_short_16","alias_value":"4P5P4J65QSIF7N3M","created_at":"2026-07-05T04:05:11.610082+00:00"},{"alias_kind":"pith_short_8","alias_value":"4P5P4J65","created_at":"2026-07-05T04:05:11.610082+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2211.15089","citing_title":"Continuous diffusion for categorical data","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG","json":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG.json","graph_json":"https://pith.science/api/pith-number/4P5P4J65QSIF7N3MCWJFFFDWTG/graph.json","events_json":"https://pith.science/api/pith-number/4P5P4J65QSIF7N3MCWJFFFDWTG/events.json","paper":"https://pith.science/paper/4P5P4J65"},"agent_actions":{"view_html":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG","download_json":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG.json","view_paper":"https://pith.science/paper/4P5P4J65","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.02736&json=true","fetch_graph":"https://pith.science/api/pith-number/4P5P4J65QSIF7N3MCWJFFFDWTG/graph.json","fetch_events":"https://pith.science/api/pith-number/4P5P4J65QSIF7N3MCWJFFFDWTG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG/action/storage_attestation","attest_author":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG/action/author_attestation","sign_citation":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG/action/citation_signature","submit_replication":"https://pith.science/pith/4P5P4J65QSIF7N3MCWJFFFDWTG/action/replication_record"}},"created_at":"2026-07-05T04:05:11.610082+00:00","updated_at":"2026-07-05T04:05:11.610082+00:00"}