{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WZIB3CVGGQLAEM7AYEZKK2ZZON","short_pith_number":"pith:WZIB3CVG","schema_version":"1.0","canonical_sha256":"b6501d8aa634160233e0c132a56b39736d99dd15bf2a3cdba7022876fd364cae","source":{"kind":"arxiv","id":"2606.08417","version":1},"attestation_state":"computed","paper":{"title":"Hacking Generative Perplexity: Why Unconditional Text Evaluation Needs Distributional Metrics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alexander Tong, Antonio Franca","submitted_at":"2026-06-07T02:35:56Z","abstract_excerpt":"Diffusion and continuous flow-based language models have emerged as the leading non-autoregressive alternatives to language modeling. Progress in both paradigms is overwhelmingly tracked by generative perplexity (gen-PPL): the per-token negative log-likelihood of samples under a frozen autoregressive (AR) scorer such as gpt2-large, typically paired with an empirical-entropy guardrail to rule out low-entropy collapse. We argue that this metric is unsound. By construction, gen-PPL measures only predictability under the scoring AR, not grammaticality or semantic coherence -- and the set of predic"},"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":"2606.08417","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-06-07T02:35:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1d58bf9f458fa2ab1a1ccf7bd092ed274cc6cbfb123d27d5a6cea257521a460f","abstract_canon_sha256":"dd7dd6bb4119b988007a86a3a30312d715d8b05adbe571d88149c7d71ac92df4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:05:36.132082Z","signature_b64":"+JPBLOTksQy1OxJ0KVtiLsML/xUU3aIFceAd99W28uPhv6F5ZmZfPqAk/utYg25kvqVoVmDO0/Qg6KlKfz8qBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b6501d8aa634160233e0c132a56b39736d99dd15bf2a3cdba7022876fd364cae","last_reissued_at":"2026-06-09T01:05:36.131538Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:05:36.131538Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hacking Generative Perplexity: Why Unconditional Text Evaluation Needs Distributional Metrics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alexander Tong, Antonio Franca","submitted_at":"2026-06-07T02:35:56Z","abstract_excerpt":"Diffusion and continuous flow-based language models have emerged as the leading non-autoregressive alternatives to language modeling. Progress in both paradigms is overwhelmingly tracked by generative perplexity (gen-PPL): the per-token negative log-likelihood of samples under a frozen autoregressive (AR) scorer such as gpt2-large, typically paired with an empirical-entropy guardrail to rule out low-entropy collapse. We argue that this metric is unsound. By construction, gen-PPL measures only predictability under the scoring AR, not grammaticality or semantic coherence -- and the set of predic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.08417","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/2606.08417/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":"2606.08417","created_at":"2026-06-09T01:05:36.131607+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.08417v1","created_at":"2026-06-09T01:05:36.131607+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.08417","created_at":"2026-06-09T01:05:36.131607+00:00"},{"alias_kind":"pith_short_12","alias_value":"WZIB3CVGGQLA","created_at":"2026-06-09T01:05:36.131607+00:00"},{"alias_kind":"pith_short_16","alias_value":"WZIB3CVGGQLAEM7A","created_at":"2026-06-09T01:05:36.131607+00:00"},{"alias_kind":"pith_short_8","alias_value":"WZIB3CVG","created_at":"2026-06-09T01:05:36.131607+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/WZIB3CVGGQLAEM7AYEZKK2ZZON","json":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON.json","graph_json":"https://pith.science/api/pith-number/WZIB3CVGGQLAEM7AYEZKK2ZZON/graph.json","events_json":"https://pith.science/api/pith-number/WZIB3CVGGQLAEM7AYEZKK2ZZON/events.json","paper":"https://pith.science/paper/WZIB3CVG"},"agent_actions":{"view_html":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON","download_json":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON.json","view_paper":"https://pith.science/paper/WZIB3CVG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.08417&json=true","fetch_graph":"https://pith.science/api/pith-number/WZIB3CVGGQLAEM7AYEZKK2ZZON/graph.json","fetch_events":"https://pith.science/api/pith-number/WZIB3CVGGQLAEM7AYEZKK2ZZON/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON/action/storage_attestation","attest_author":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON/action/author_attestation","sign_citation":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON/action/citation_signature","submit_replication":"https://pith.science/pith/WZIB3CVGGQLAEM7AYEZKK2ZZON/action/replication_record"}},"created_at":"2026-06-09T01:05:36.131607+00:00","updated_at":"2026-06-09T01:05:36.131607+00:00"}