{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:VDK6MBX55VGO2B46SXUS5ESXZY","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"bc1667a62d044b06d87094c993236f37b2668814313f48ae7ffeead1c276eec0","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-16T06:33:38Z","title_canon_sha256":"e824f1a709c8e7b834f17d7af1268258aae06bbd1521fd7f212830335d200e83"},"schema_version":"1.0","source":{"id":"2605.16834","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.16834","created_at":"2026-05-20T00:03:25Z"},{"alias_kind":"arxiv_version","alias_value":"2605.16834v1","created_at":"2026-05-20T00:03:25Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16834","created_at":"2026-05-20T00:03:25Z"},{"alias_kind":"pith_short_12","alias_value":"VDK6MBX55VGO","created_at":"2026-05-20T00:03:25Z"},{"alias_kind":"pith_short_16","alias_value":"VDK6MBX55VGO2B46","created_at":"2026-05-20T00:03:25Z"},{"alias_kind":"pith_short_8","alias_value":"VDK6MBX5","created_at":"2026-05-20T00:03:25Z"}],"graph_snapshots":[{"event_id":"sha256:eb3bc4f3b4674cd30e6afd4e8ab3f83a7e0d87259781bb203b325194b9784821","target":"graph","created_at":"2026-05-20T00:03:25Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Despite learning only the anchors without heavy projection layers, our approach consistently outperforms existing methods in zero-shot classification, cross-modal retrieval, and zero-shot segmentation by a substantial margin."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That training a set of learnable anchors to induce consistent cross-modal similarity patterns for matched pairs is sufficient to capture fine-grained token-level relations without requiring additional projection layers or larger paired datasets."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"A new post-hoc alignment technique uses learnable anchors to capture token-level relative similarities between modalities, outperforming global alignment baselines on zero-shot classification, retrieval, and segmentation with scarce paired examples."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Relative representations via learnable anchors align token-level structures across modalities using only limited paired examples."}],"snapshot_sha256":"777d2eef5a35ef6d232a757a47b54f8b3170a4f641b64bcc37565a76bdddecca"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"ddf4b06482223e22f26d8fa6ef40834b6630c1fc4b8b0375c5854928715e4f0d"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T21:01:19.248521Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T20:51:12.437387Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T19:01:56.254703Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.398949Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.16834/integrity.json","findings":[],"snapshot_sha256":"e3c8db28b098ec01440f520c3b6850965316f6b9e479d244961270d94a7a134f","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Multimodal pre-training demonstrates strong generalization performance, but this paradigm is often impractical in domains where paired data are scarce. A promising alternative is post-hoc multimodal alignment, which aligns separately pre-trained unimodal encoders using a limited number of paired examples. However, existing methods focus primarily on aligning global representations, missing patch-token relations. This may hinder transfer to tasks that require fine-grained cross-modal matching beyond coarse sample-level semantics. To address this issue, we propose a post-hoc alignment method tha","authors_text":"Shiwon Kim, Yu Rang Park","cross_cats":["cs.AI","cs.LG"],"headline":"Relative representations via learnable anchors align token-level structures across modalities using only limited paired examples.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-16T06:33:38Z","title":"Learning Relative Representations for Fine-Grained Multimodal Alignment with Limited Data"},"references":{"count":31,"internal_anchors":2,"resolved_work":31,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Learning transferable visual models from natural language supervision","work_id":"9ff70d47-d644-4c21-a8f2-2e01a0734b97","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Le, Yun- Hsuan Sung, Zhen Li, and Tom Duerig","work_id":"8d222499-da6d-44c6-80d6-d4490667ce96","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"The platonic representation hypothesis","work_id":"7e23235a-add0-43c4-8e54-bb2b253439c8","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Relative representations enable zero-shot latent space communication","work_id":"1ce4feb0-b01e-4d77-93f4-86427be53c00","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Linearly mapping from image to text space.arXiv preprint arXiv:2209.15162","work_id":"1470dc2c-e9da-4e3f-aa62-808e9aa7bc2e","year":2022}],"snapshot_sha256":"08c7405dc4cdc8623b7973f217df8c480ff121cf8c44e3cdffa9195f18d58927"},"source":{"id":"2605.16834","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T20:45:06.663156Z","id":"6033a144-e12e-4919-a6e6-7670c8ff92b3","model_set":{"reader":"grok-4.3"},"one_line_summary":"A new post-hoc alignment technique uses learnable anchors to capture token-level relative similarities between modalities, outperforming global alignment baselines on zero-shot classification, retrieval, and segmentation with scarce paired examples.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Relative representations via learnable anchors align token-level structures across modalities using only limited paired examples.","strongest_claim":"Despite learning only the anchors without heavy projection layers, our approach consistently outperforms existing methods in zero-shot classification, cross-modal retrieval, and zero-shot segmentation by a substantial margin.","weakest_assumption":"That training a set of learnable anchors to induce consistent cross-modal similarity patterns for matched pairs is sufficient to capture fine-grained token-level relations without requiring additional projection layers or larger paired datasets."}},"verdict_id":"6033a144-e12e-4919-a6e6-7670c8ff92b3"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:361b39d299099d9cd4954a2cfe1d023ff9d5e7e84952a879fae8367c5f6586cf","target":"record","created_at":"2026-05-20T00:03:25Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"bc1667a62d044b06d87094c993236f37b2668814313f48ae7ffeead1c276eec0","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-16T06:33:38Z","title_canon_sha256":"e824f1a709c8e7b834f17d7af1268258aae06bbd1521fd7f212830335d200e83"},"schema_version":"1.0","source":{"id":"2605.16834","kind":"arxiv","version":1}},"canonical_sha256":"a8d5e606fded4ced079e95e92e9257ce3c77e92056f481849826a2b782aa5116","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a8d5e606fded4ced079e95e92e9257ce3c77e92056f481849826a2b782aa5116","first_computed_at":"2026-05-20T00:03:25.229675Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:25.229675Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"CKhUrcI3D8YC4YuhVAjyNg3iLboRhZTCuaeU6IwAV7SoulbYFQda7SZnRFXTzJt4g7dls1h2S+8V4kbkOJ0yCg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:25.230505Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.16834","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:361b39d299099d9cd4954a2cfe1d023ff9d5e7e84952a879fae8367c5f6586cf","sha256:eb3bc4f3b4674cd30e6afd4e8ab3f83a7e0d87259781bb203b325194b9784821"],"state_sha256":"a03f0becdbf0d070ae17c904f6b79437ecd6859b04181249a10bd9180cc50905"}