{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:PIRWNFJW22DFCAMEG2DF7TWZL3","short_pith_number":"pith:PIRWNFJW","schema_version":"1.0","canonical_sha256":"7a23669536d68651018436865fced95ec0741e5db15cae18fe7d4eda1846be48","source":{"kind":"arxiv","id":"2302.14383","version":3},"attestation_state":"computed","paper":{"title":"Linear Spaces of Meanings: Compositional Structures in Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Alessandro Achille, Luca Zancato, Matthew Trager, Parminder Bhatia, Pramuditha Perera, Stefano Soatto","submitted_at":"2023-02-28T08:11:56Z","abstract_excerpt":"We investigate compositional structures in data embeddings from pre-trained vision-language models (VLMs). Traditionally, compositionality has been associated with algebraic operations on embeddings of words from a pre-existing vocabulary. In contrast, we seek to approximate representations from an encoder as combinations of a smaller set of vectors in the embedding space. These vectors can be seen as \"ideal words\" for generating concepts directly within the embedding space of the model. We first present a framework for understanding compositional structures from a geometric perspective. We th"},"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":"2302.14383","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-28T08:11:56Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"title_canon_sha256":"8db547ef1fe08057c603eeca147b9197b7d47293216365bf1ca54cb90366e2c6","abstract_canon_sha256":"f75be5da4d3e045646b42d53c3a92c1b2f29952bdc5f1c13b3c6880eacbabc98"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:32:17.164742Z","signature_b64":"viuWAaJYGJng9N5NQjh5HvqGz91dWmstCzsxkcs2b8ziAolc7iVNUpP+xFLL+lQL94cJKBMeVaZSeez3RxSKCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7a23669536d68651018436865fced95ec0741e5db15cae18fe7d4eda1846be48","last_reissued_at":"2026-07-05T07:32:17.164240Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:32:17.164240Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Linear Spaces of Meanings: Compositional Structures in Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Alessandro Achille, Luca Zancato, Matthew Trager, Parminder Bhatia, Pramuditha Perera, Stefano Soatto","submitted_at":"2023-02-28T08:11:56Z","abstract_excerpt":"We investigate compositional structures in data embeddings from pre-trained vision-language models (VLMs). Traditionally, compositionality has been associated with algebraic operations on embeddings of words from a pre-existing vocabulary. In contrast, we seek to approximate representations from an encoder as combinations of a smaller set of vectors in the embedding space. These vectors can be seen as \"ideal words\" for generating concepts directly within the embedding space of the model. We first present a framework for understanding compositional structures from a geometric perspective. We th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.14383","kind":"arxiv","version":3},"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/2302.14383/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":"2302.14383","created_at":"2026-07-05T07:32:17.164296+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.14383v3","created_at":"2026-07-05T07:32:17.164296+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.14383","created_at":"2026-07-05T07:32:17.164296+00:00"},{"alias_kind":"pith_short_12","alias_value":"PIRWNFJW22DF","created_at":"2026-07-05T07:32:17.164296+00:00"},{"alias_kind":"pith_short_16","alias_value":"PIRWNFJW22DFCAME","created_at":"2026-07-05T07:32:17.164296+00:00"},{"alias_kind":"pith_short_8","alias_value":"PIRWNFJW","created_at":"2026-07-05T07:32:17.164296+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.03340","citing_title":"Learning Additively Compositional Latent Actions for Embodied AI","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11203","citing_title":"FeatMap: Understanding image manipulation in the feature space and its implications for feature space geometry","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3","json":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3.json","graph_json":"https://pith.science/api/pith-number/PIRWNFJW22DFCAMEG2DF7TWZL3/graph.json","events_json":"https://pith.science/api/pith-number/PIRWNFJW22DFCAMEG2DF7TWZL3/events.json","paper":"https://pith.science/paper/PIRWNFJW"},"agent_actions":{"view_html":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3","download_json":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3.json","view_paper":"https://pith.science/paper/PIRWNFJW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.14383&json=true","fetch_graph":"https://pith.science/api/pith-number/PIRWNFJW22DFCAMEG2DF7TWZL3/graph.json","fetch_events":"https://pith.science/api/pith-number/PIRWNFJW22DFCAMEG2DF7TWZL3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3/action/storage_attestation","attest_author":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3/action/author_attestation","sign_citation":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3/action/citation_signature","submit_replication":"https://pith.science/pith/PIRWNFJW22DFCAMEG2DF7TWZL3/action/replication_record"}},"created_at":"2026-07-05T07:32:17.164296+00:00","updated_at":"2026-07-05T07:32:17.164296+00:00"}