{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CSLCRUY3HWBKEQ7AAQBNBQ3LMB","short_pith_number":"pith:CSLCRUY3","schema_version":"1.0","canonical_sha256":"149628d31b3d82a243e00402d0c36b606e58b719748fd20ff89ac7f42a659bdf","source":{"kind":"arxiv","id":"2410.11087","version":2},"attestation_state":"computed","paper":{"title":"Locality Alignment Improves Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ian Covert, James Zou, Tatsunori Hashimoto, Tony Sun","submitted_at":"2024-10-14T21:01:01Z","abstract_excerpt":"Vision language models (VLMs) have seen growing adoption in recent years, but many still struggle with basic spatial reasoning errors. We hypothesize that this is due to VLMs adopting pre-trained vision backbones, specifically vision transformers (ViTs) trained with image-level supervision and minimal inductive biases. Such models may fail to encode the class contents at each position in the image, and our goal is to resolve this with a vision backbone that effectively captures both local and global image semantics. Our main insight is that we do not require new supervision to learn this capab"},"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":"2410.11087","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-10-14T21:01:01Z","cross_cats_sorted":[],"title_canon_sha256":"301f41507517cb0460f1ec12d189bc29879b484a299b2786ab0f410c183e4c55","abstract_canon_sha256":"63e8d20d32238d6f0a15765cfd5e9a18c764d66d2833ee34963415ec62d5ce30"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:23:24.125219Z","signature_b64":"5ZSHt7IAbnmgaDl26z/VXhc4VJbhH49UW7r4peAz4LaOpmyH2/jKdeH+tyBBLAlenDWJR4Ntj3OQuFfTBxQXCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"149628d31b3d82a243e00402d0c36b606e58b719748fd20ff89ac7f42a659bdf","last_reissued_at":"2026-07-05T10:23:24.124308Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:23:24.124308Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Locality Alignment Improves Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ian Covert, James Zou, Tatsunori Hashimoto, Tony Sun","submitted_at":"2024-10-14T21:01:01Z","abstract_excerpt":"Vision language models (VLMs) have seen growing adoption in recent years, but many still struggle with basic spatial reasoning errors. We hypothesize that this is due to VLMs adopting pre-trained vision backbones, specifically vision transformers (ViTs) trained with image-level supervision and minimal inductive biases. Such models may fail to encode the class contents at each position in the image, and our goal is to resolve this with a vision backbone that effectively captures both local and global image semantics. Our main insight is that we do not require new supervision to learn this capab"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.11087","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/2410.11087/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":"2410.11087","created_at":"2026-07-05T10:23:24.124471+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.11087v2","created_at":"2026-07-05T10:23:24.124471+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.11087","created_at":"2026-07-05T10:23:24.124471+00:00"},{"alias_kind":"pith_short_12","alias_value":"CSLCRUY3HWBK","created_at":"2026-07-05T10:23:24.124471+00:00"},{"alias_kind":"pith_short_16","alias_value":"CSLCRUY3HWBKEQ7A","created_at":"2026-07-05T10:23:24.124471+00:00"},{"alias_kind":"pith_short_8","alias_value":"CSLCRUY3","created_at":"2026-07-05T10:23:24.124471+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07907","citing_title":"Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarks","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28401","citing_title":"Vision-driven Preference Synthesis for Mitigating Hallucinations in VLMs","ref_index":99,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21343","citing_title":"Latent Denoising Improves Visual Alignment in Large Multimodal Models","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB","json":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB.json","graph_json":"https://pith.science/api/pith-number/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/graph.json","events_json":"https://pith.science/api/pith-number/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/events.json","paper":"https://pith.science/paper/CSLCRUY3"},"agent_actions":{"view_html":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB","download_json":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB.json","view_paper":"https://pith.science/paper/CSLCRUY3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.11087&json=true","fetch_graph":"https://pith.science/api/pith-number/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/graph.json","fetch_events":"https://pith.science/api/pith-number/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/action/storage_attestation","attest_author":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/action/author_attestation","sign_citation":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/action/citation_signature","submit_replication":"https://pith.science/pith/CSLCRUY3HWBKEQ7AAQBNBQ3LMB/action/replication_record"}},"created_at":"2026-07-05T10:23:24.124471+00:00","updated_at":"2026-07-05T10:23:24.124471+00:00"}