{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IFWDKKLGLFU5KMOKJPIRAZG5OH","short_pith_number":"pith:IFWDKKLG","schema_version":"1.0","canonical_sha256":"416c3529665969d531ca4bd11064dd71d956b3b240909c2204ed7e182e05df8d","source":{"kind":"arxiv","id":"2503.12355","version":1},"attestation_state":"computed","paper":{"title":"Atlas: Multi-Scale Attention Improves Long Context Image Modeling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Adam Yala, Alexander Bick, Boyi Li, Kumar Krishna Agrawal, Longchao Liu, Long Lian, Maggie Chung, Natalia Harguindeguy, Trevor Darrell","submitted_at":"2025-03-16T04:52:13Z","abstract_excerpt":"Efficiently modeling massive images is a long-standing challenge in machine learning. To this end, we introduce Multi-Scale Attention (MSA). MSA relies on two key ideas, (i) multi-scale representations (ii) bi-directional cross-scale communication. MSA creates O(log N) scales to represent the image across progressively coarser features and leverages cross-attention to propagate information across scales. We then introduce Atlas, a novel neural network architecture based on MSA. We demonstrate that Atlas significantly improves the compute-performance tradeoff of long-context image modeling in a"},"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":"2503.12355","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-16T04:52:13Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"df2c8ef22761820c8eae5c2c60f51b7ddbf7a5410f3322680d8d868159031830","abstract_canon_sha256":"a65c9667ec0f47a59221d5384c96fde450d829004911fb420e145556ce38eed8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:32:11.146581Z","signature_b64":"/HpBaxHWHFcImRa/AwdI+YXGJj5hWMod295EHW05d6dKbe73FGt6Gl+YRPIP4hZr0y1AVOdqYIEWJXgoPwrlCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"416c3529665969d531ca4bd11064dd71d956b3b240909c2204ed7e182e05df8d","last_reissued_at":"2026-07-05T10:32:11.145860Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:32:11.145860Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Atlas: Multi-Scale Attention Improves Long Context Image Modeling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Adam Yala, Alexander Bick, Boyi Li, Kumar Krishna Agrawal, Longchao Liu, Long Lian, Maggie Chung, Natalia Harguindeguy, Trevor Darrell","submitted_at":"2025-03-16T04:52:13Z","abstract_excerpt":"Efficiently modeling massive images is a long-standing challenge in machine learning. To this end, we introduce Multi-Scale Attention (MSA). MSA relies on two key ideas, (i) multi-scale representations (ii) bi-directional cross-scale communication. MSA creates O(log N) scales to represent the image across progressively coarser features and leverages cross-attention to propagate information across scales. We then introduce Atlas, a novel neural network architecture based on MSA. We demonstrate that Atlas significantly improves the compute-performance tradeoff of long-context image modeling in a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.12355","kind":"arxiv","version":1},"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/2503.12355/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":"2503.12355","created_at":"2026-07-05T10:32:11.145946+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.12355v1","created_at":"2026-07-05T10:32:11.145946+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.12355","created_at":"2026-07-05T10:32:11.145946+00:00"},{"alias_kind":"pith_short_12","alias_value":"IFWDKKLGLFU5","created_at":"2026-07-05T10:32:11.145946+00:00"},{"alias_kind":"pith_short_16","alias_value":"IFWDKKLGLFU5KMOK","created_at":"2026-07-05T10:32:11.145946+00:00"},{"alias_kind":"pith_short_8","alias_value":"IFWDKKLG","created_at":"2026-07-05T10:32:11.145946+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24570","citing_title":"Jolia: Concept-Level Vision-Language Alignment for 3D CT Contrastive Learning","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17140","citing_title":"UCSF-PDGM-VQA: Visual Question Answering dataset for brain tumor MRI interpretation","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17140","citing_title":"UCSF-PDGM-VQA: Visual Question Answering dataset for brain tumor MRI interpretation","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10437","citing_title":"Enhancing Fine-Grained Spatial Grounding in 3D CT Report Generation via Discriminative Guidance","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH","json":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH.json","graph_json":"https://pith.science/api/pith-number/IFWDKKLGLFU5KMOKJPIRAZG5OH/graph.json","events_json":"https://pith.science/api/pith-number/IFWDKKLGLFU5KMOKJPIRAZG5OH/events.json","paper":"https://pith.science/paper/IFWDKKLG"},"agent_actions":{"view_html":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH","download_json":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH.json","view_paper":"https://pith.science/paper/IFWDKKLG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.12355&json=true","fetch_graph":"https://pith.science/api/pith-number/IFWDKKLGLFU5KMOKJPIRAZG5OH/graph.json","fetch_events":"https://pith.science/api/pith-number/IFWDKKLGLFU5KMOKJPIRAZG5OH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH/action/storage_attestation","attest_author":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH/action/author_attestation","sign_citation":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH/action/citation_signature","submit_replication":"https://pith.science/pith/IFWDKKLGLFU5KMOKJPIRAZG5OH/action/replication_record"}},"created_at":"2026-07-05T10:32:11.145946+00:00","updated_at":"2026-07-05T10:32:11.145946+00:00"}