{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DHKIQ34W4WY5IYCJLGCU2LVWRL","short_pith_number":"pith:DHKIQ34W","schema_version":"1.0","canonical_sha256":"19d4886f96e5b1d4604959854d2eb68acfb775348629f7094d543075a8185ada","source":{"kind":"arxiv","id":"2507.20956","version":1},"attestation_state":"computed","paper":{"title":"Mind the Gap: Conformative Decoding to Improve Output Diversity of Instruction-Tuned Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Anna Jordanous, Dan Brown, Max Peeperkorn, Tom Kouwenhoven","submitted_at":"2025-07-28T16:04:25Z","abstract_excerpt":"Instruction-tuning large language models (LLMs) reduces the diversity of their outputs, which has implications for many tasks, particularly for creative tasks. This paper investigates the ``diversity gap'' for a writing prompt narrative generation task. This gap emerges as measured by current diversity metrics for various open-weight and open-source LLMs. The results show significant decreases in diversity due to instruction-tuning. We explore the diversity loss at each fine-tuning stage for the OLMo and OLMo 2 models to further understand how output diversity is affected. The results indicate"},"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":"2507.20956","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-28T16:04:25Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1fd0ac80d8d56520da38ff4d4b28bf533c9b42d351c799f983dffd56552d509a","abstract_canon_sha256":"fff9173da42171be4afe9804813faafeb0d8c79d38fc1a1f877b5cd6a084b67c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:44:29.013887Z","signature_b64":"I95cDPrzDXkj4K4Te+9DGLjxEMuc+AYDKSlRjiW1giqa6665GcXCtzd/7weT0c6KrSWHoINDGo765OZKn7WhBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"19d4886f96e5b1d4604959854d2eb68acfb775348629f7094d543075a8185ada","last_reissued_at":"2026-07-05T11:44:29.013417Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:44:29.013417Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mind the Gap: Conformative Decoding to Improve Output Diversity of Instruction-Tuned Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Anna Jordanous, Dan Brown, Max Peeperkorn, Tom Kouwenhoven","submitted_at":"2025-07-28T16:04:25Z","abstract_excerpt":"Instruction-tuning large language models (LLMs) reduces the diversity of their outputs, which has implications for many tasks, particularly for creative tasks. This paper investigates the ``diversity gap'' for a writing prompt narrative generation task. This gap emerges as measured by current diversity metrics for various open-weight and open-source LLMs. The results show significant decreases in diversity due to instruction-tuning. We explore the diversity loss at each fine-tuning stage for the OLMo and OLMo 2 models to further understand how output diversity is affected. The results indicate"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.20956","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/2507.20956/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":"2507.20956","created_at":"2026-07-05T11:44:29.013474+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.20956v1","created_at":"2026-07-05T11:44:29.013474+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.20956","created_at":"2026-07-05T11:44:29.013474+00:00"},{"alias_kind":"pith_short_12","alias_value":"DHKIQ34W4WY5","created_at":"2026-07-05T11:44:29.013474+00:00"},{"alias_kind":"pith_short_16","alias_value":"DHKIQ34W4WY5IYCJ","created_at":"2026-07-05T11:44:29.013474+00:00"},{"alias_kind":"pith_short_8","alias_value":"DHKIQ34W","created_at":"2026-07-05T11:44:29.013474+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12790","citing_title":"GENIE: A Fine-Grained Measure for Novelty","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2601.06116","citing_title":"The Homogenization Problem in LLMs: Towards Meaningful Diversity in AI Safety","ref_index":98,"is_internal_anchor":false},{"citing_arxiv_id":"2601.06116","citing_title":"The Homogenization Problem in LLMs: Towards Meaningful Diversity in AI Safety","ref_index":98,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL","json":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL.json","graph_json":"https://pith.science/api/pith-number/DHKIQ34W4WY5IYCJLGCU2LVWRL/graph.json","events_json":"https://pith.science/api/pith-number/DHKIQ34W4WY5IYCJLGCU2LVWRL/events.json","paper":"https://pith.science/paper/DHKIQ34W"},"agent_actions":{"view_html":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL","download_json":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL.json","view_paper":"https://pith.science/paper/DHKIQ34W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.20956&json=true","fetch_graph":"https://pith.science/api/pith-number/DHKIQ34W4WY5IYCJLGCU2LVWRL/graph.json","fetch_events":"https://pith.science/api/pith-number/DHKIQ34W4WY5IYCJLGCU2LVWRL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL/action/storage_attestation","attest_author":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL/action/author_attestation","sign_citation":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL/action/citation_signature","submit_replication":"https://pith.science/pith/DHKIQ34W4WY5IYCJLGCU2LVWRL/action/replication_record"}},"created_at":"2026-07-05T11:44:29.013474+00:00","updated_at":"2026-07-05T11:44:29.013474+00:00"}