{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:B2V5EDZPOKB6JPOHEXZ3ETRMEG","short_pith_number":"pith:B2V5EDZP","schema_version":"1.0","canonical_sha256":"0eabd20f2f7283e4bdc725f3b24e2c219fcac3374cac0a9c7fa1daf4884deb7c","source":{"kind":"arxiv","id":"2406.09358","version":2},"attestation_state":"computed","paper":{"title":"Understanding Hallucinations in Diffusion Models through Mode Interpolation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"J. Zico Kolter, Pratyush Maini, Sumukh K Aithal, Zachary C. Lipton","submitted_at":"2024-06-13T17:43:41Z","abstract_excerpt":"Colloquially speaking, image generation models based upon diffusion processes are frequently said to exhibit \"hallucinations,\" samples that could never occur in the training data. But where do such hallucinations come from? In this paper, we study a particular failure mode in diffusion models, which we term mode interpolation. Specifically, we find that diffusion models smoothly \"interpolate\" between nearby data modes in the training set, to generate samples that are completely outside the support of the original training distribution; this phenomenon leads diffusion models to generate artifac"},"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":"2406.09358","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-13T17:43:41Z","cross_cats_sorted":[],"title_canon_sha256":"22b42c34a7d4e3c1cbab6aea11b3bf4b170ea27bd3f82e187e59b0ef062d6ac7","abstract_canon_sha256":"e54bb86a9e1e1f0829b2f9d1b1f3adcd299c6005b27d0a960784a56cb1f0d97e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:59:10.055112Z","signature_b64":"QOL2ZbYKNaMm/scotfddzjUqPMNTRG1REoy3IqBnqgPQJ5RqtrBcw2smgCV3DJOccD2mdwfXrknOdzOhf9zjDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0eabd20f2f7283e4bdc725f3b24e2c219fcac3374cac0a9c7fa1daf4884deb7c","last_reissued_at":"2026-07-05T08:59:10.054613Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:59:10.054613Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding Hallucinations in Diffusion Models through Mode Interpolation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"J. Zico Kolter, Pratyush Maini, Sumukh K Aithal, Zachary C. Lipton","submitted_at":"2024-06-13T17:43:41Z","abstract_excerpt":"Colloquially speaking, image generation models based upon diffusion processes are frequently said to exhibit \"hallucinations,\" samples that could never occur in the training data. But where do such hallucinations come from? In this paper, we study a particular failure mode in diffusion models, which we term mode interpolation. Specifically, we find that diffusion models smoothly \"interpolate\" between nearby data modes in the training set, to generate samples that are completely outside the support of the original training distribution; this phenomenon leads diffusion models to generate artifac"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.09358","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/2406.09358/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":"2406.09358","created_at":"2026-07-05T08:59:10.054673+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.09358v2","created_at":"2026-07-05T08:59:10.054673+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.09358","created_at":"2026-07-05T08:59:10.054673+00:00"},{"alias_kind":"pith_short_12","alias_value":"B2V5EDZPOKB6","created_at":"2026-07-05T08:59:10.054673+00:00"},{"alias_kind":"pith_short_16","alias_value":"B2V5EDZPOKB6JPOH","created_at":"2026-07-05T08:59:10.054673+00:00"},{"alias_kind":"pith_short_8","alias_value":"B2V5EDZP","created_at":"2026-07-05T08:59:10.054673+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2601.06116","citing_title":"The Homogenization Problem in LLMs: Towards Meaningful Diversity in AI Safety","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20299","citing_title":"Mechanisms of Misgeneralization in Physical Sequence Modeling","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16520","citing_title":"Global Convergence of Sampling-Based Nonconvex Optimization through Diffusion-Style Smoothing","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2506.06024","citing_title":"On Inverse Problems, Parameter Estimation, and Domain Generalization","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2601.06116","citing_title":"The Homogenization Problem in LLMs: Towards Meaningful Diversity in AI Safety","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2410.05229","citing_title":"GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models","ref_index":33,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG","json":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG.json","graph_json":"https://pith.science/api/pith-number/B2V5EDZPOKB6JPOHEXZ3ETRMEG/graph.json","events_json":"https://pith.science/api/pith-number/B2V5EDZPOKB6JPOHEXZ3ETRMEG/events.json","paper":"https://pith.science/paper/B2V5EDZP"},"agent_actions":{"view_html":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG","download_json":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG.json","view_paper":"https://pith.science/paper/B2V5EDZP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.09358&json=true","fetch_graph":"https://pith.science/api/pith-number/B2V5EDZPOKB6JPOHEXZ3ETRMEG/graph.json","fetch_events":"https://pith.science/api/pith-number/B2V5EDZPOKB6JPOHEXZ3ETRMEG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG/action/storage_attestation","attest_author":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG/action/author_attestation","sign_citation":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG/action/citation_signature","submit_replication":"https://pith.science/pith/B2V5EDZPOKB6JPOHEXZ3ETRMEG/action/replication_record"}},"created_at":"2026-07-05T08:59:10.054673+00:00","updated_at":"2026-07-05T08:59:10.054673+00:00"}