{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2H6WV2KWI5PXPE5O7UFVDEURXM","short_pith_number":"pith:2H6WV2KW","schema_version":"1.0","canonical_sha256":"d1fd6ae956475f7793aefd0b519291bb0d1c60cca20043265dc7c238b834f221","source":{"kind":"arxiv","id":"2510.04533","version":2},"attestation_state":"computed","paper":{"title":"TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Donghoon Ahn, Hyunmin Cho, Jee Eun Kim, Kyong Hwan Jin, Seungryong Kim, Susung Hong","submitted_at":"2025-10-06T06:53:29Z","abstract_excerpt":"Diffusion models achieve state-of-the-art image generation but often produce semantic inconsistencies, or hallucinations. Existing inference-time guidance methods rely on external signals or architectural modifications, adding computational overhead. We propose $\\mathbf{T}$angential $\\mathbf{A}$mplifying $\\mathbf{G}$uidance $\\mathbf{(TAG)}$, a training-free, architecture-agnostic, plug-and-play guidance method that operates purely on trajectory signals. TAG uses an intermediate sample as a projection basis and amplifies the tangential components of the estimated score to correct the sampling t"},"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":"2510.04533","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-10-06T06:53:29Z","cross_cats_sorted":[],"title_canon_sha256":"7cdb858823c5c3282cf15b95891b99058b934e95d7f226f7690681d52d6122c0","abstract_canon_sha256":"524b3e8b6f3ba1eb6d178848f7c4d8d579148578275d76ea3874981cf84f1e70"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:05:38.402128Z","signature_b64":"AGlQC9OBZw2SQh4bV1KlOLz/4vpywPo6M08S/qCSCPcHtWFrMOtJ4OEAND5SEqOa9LsBgDLf9CsGfKT2G8gUBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d1fd6ae956475f7793aefd0b519291bb0d1c60cca20043265dc7c238b834f221","last_reissued_at":"2026-05-27T01:05:38.401253Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:05:38.401253Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Donghoon Ahn, Hyunmin Cho, Jee Eun Kim, Kyong Hwan Jin, Seungryong Kim, Susung Hong","submitted_at":"2025-10-06T06:53:29Z","abstract_excerpt":"Diffusion models achieve state-of-the-art image generation but often produce semantic inconsistencies, or hallucinations. Existing inference-time guidance methods rely on external signals or architectural modifications, adding computational overhead. We propose $\\mathbf{T}$angential $\\mathbf{A}$mplifying $\\mathbf{G}$uidance $\\mathbf{(TAG)}$, a training-free, architecture-agnostic, plug-and-play guidance method that operates purely on trajectory signals. TAG uses an intermediate sample as a projection basis and amplifies the tangential components of the estimated score to correct the sampling t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.04533","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/2510.04533/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":"2510.04533","created_at":"2026-05-27T01:05:38.401384+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.04533v2","created_at":"2026-05-27T01:05:38.401384+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.04533","created_at":"2026-05-27T01:05:38.401384+00:00"},{"alias_kind":"pith_short_12","alias_value":"2H6WV2KWI5PX","created_at":"2026-05-27T01:05:38.401384+00:00"},{"alias_kind":"pith_short_16","alias_value":"2H6WV2KWI5PXPE5O","created_at":"2026-05-27T01:05:38.401384+00:00"},{"alias_kind":"pith_short_8","alias_value":"2H6WV2KW","created_at":"2026-05-27T01:05:38.401384+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.20079","citing_title":"Probability-Conserving Flow Guidance","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2604.08048","citing_title":"Guiding a Diffusion Model by Swapping Its Tokens","ref_index":4,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM","json":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM.json","graph_json":"https://pith.science/api/pith-number/2H6WV2KWI5PXPE5O7UFVDEURXM/graph.json","events_json":"https://pith.science/api/pith-number/2H6WV2KWI5PXPE5O7UFVDEURXM/events.json","paper":"https://pith.science/paper/2H6WV2KW"},"agent_actions":{"view_html":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM","download_json":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM.json","view_paper":"https://pith.science/paper/2H6WV2KW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.04533&json=true","fetch_graph":"https://pith.science/api/pith-number/2H6WV2KWI5PXPE5O7UFVDEURXM/graph.json","fetch_events":"https://pith.science/api/pith-number/2H6WV2KWI5PXPE5O7UFVDEURXM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM/action/storage_attestation","attest_author":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM/action/author_attestation","sign_citation":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM/action/citation_signature","submit_replication":"https://pith.science/pith/2H6WV2KWI5PXPE5O7UFVDEURXM/action/replication_record"}},"created_at":"2026-05-27T01:05:38.401384+00:00","updated_at":"2026-05-27T01:05:38.401384+00:00"}