{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FSCADLXOZ5ZRR3APECRBV7SDMJ","short_pith_number":"pith:FSCADLXO","schema_version":"1.0","canonical_sha256":"2c8401aeeecf7318ec0f20a21afe43625072a781ac4cffd39503b937ea5d9c0a","source":{"kind":"arxiv","id":"2405.05751","version":3},"attestation_state":"computed","paper":{"title":"Mirage: A Multi-Level Superoptimizer for Tensor Programs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.PL"],"primary_cat":"cs.LG","authors_text":"Chunan Shi, Jianan Ji, Kit Ao, Mengdi Wu, Oded Padon, Praveen Velliengiri, Shengyu Liu, Xinhao Cheng, Xupeng Miao, Zhihao Jia","submitted_at":"2024-05-09T13:15:40Z","abstract_excerpt":"We introduce Mirage, the first multi-level superoptimizer for tensor programs. A key idea in Mirage is $\\mu$Graphs, a uniform representation of tensor programs at the kernel, thread block, and thread levels of the GPU compute hierarchy. $\\mu$Graphs enable Mirage to discover novel optimizations that combine algebraic transformations, schedule transformations, and generation of new custom kernels. To navigate the large search space, Mirage introduces a pruning technique based on abstraction that significantly reduces the search space and provides a certain optimality guarantee. To ensure that 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":"2405.05751","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-09T13:15:40Z","cross_cats_sorted":["cs.AI","cs.PL"],"title_canon_sha256":"940db18b58c208fb0d98b637516f71e5f8f17e14100cdc4a049af2700202126a","abstract_canon_sha256":"10e1d1201e7b49d8a5c817b042e3e6f56a7b2dadd5eeae4acb7eb4396eb4473a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:49.072684Z","signature_b64":"Q0jWzT8YpY6HP92naztGbc7dFPPw+C0NHsb/vVgXOVhzEKw+b93kp/sZARmklfQPS8Qbl+nnIJJh0A06LltRBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2c8401aeeecf7318ec0f20a21afe43625072a781ac4cffd39503b937ea5d9c0a","last_reissued_at":"2026-07-05T11:16:49.072184Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:49.072184Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mirage: A Multi-Level Superoptimizer for Tensor Programs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.PL"],"primary_cat":"cs.LG","authors_text":"Chunan Shi, Jianan Ji, Kit Ao, Mengdi Wu, Oded Padon, Praveen Velliengiri, Shengyu Liu, Xinhao Cheng, Xupeng Miao, Zhihao Jia","submitted_at":"2024-05-09T13:15:40Z","abstract_excerpt":"We introduce Mirage, the first multi-level superoptimizer for tensor programs. A key idea in Mirage is $\\mu$Graphs, a uniform representation of tensor programs at the kernel, thread block, and thread levels of the GPU compute hierarchy. $\\mu$Graphs enable Mirage to discover novel optimizations that combine algebraic transformations, schedule transformations, and generation of new custom kernels. To navigate the large search space, Mirage introduces a pruning technique based on abstraction that significantly reduces the search space and provides a certain optimality guarantee. To ensure that th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.05751","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/2405.05751/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":"2405.05751","created_at":"2026-07-05T11:16:49.072242+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.05751v3","created_at":"2026-07-05T11:16:49.072242+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.05751","created_at":"2026-07-05T11:16:49.072242+00:00"},{"alias_kind":"pith_short_12","alias_value":"FSCADLXOZ5ZR","created_at":"2026-07-05T11:16:49.072242+00:00"},{"alias_kind":"pith_short_16","alias_value":"FSCADLXOZ5ZRR3AP","created_at":"2026-07-05T11:16:49.072242+00:00"},{"alias_kind":"pith_short_8","alias_value":"FSCADLXO","created_at":"2026-07-05T11:16:49.072242+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2412.05496","citing_title":"Flex Attention: A Programming Model for Generating Optimized Attention Kernels","ref_index":50,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ","json":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ.json","graph_json":"https://pith.science/api/pith-number/FSCADLXOZ5ZRR3APECRBV7SDMJ/graph.json","events_json":"https://pith.science/api/pith-number/FSCADLXOZ5ZRR3APECRBV7SDMJ/events.json","paper":"https://pith.science/paper/FSCADLXO"},"agent_actions":{"view_html":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ","download_json":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ.json","view_paper":"https://pith.science/paper/FSCADLXO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.05751&json=true","fetch_graph":"https://pith.science/api/pith-number/FSCADLXOZ5ZRR3APECRBV7SDMJ/graph.json","fetch_events":"https://pith.science/api/pith-number/FSCADLXOZ5ZRR3APECRBV7SDMJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ/action/storage_attestation","attest_author":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ/action/author_attestation","sign_citation":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ/action/citation_signature","submit_replication":"https://pith.science/pith/FSCADLXOZ5ZRR3APECRBV7SDMJ/action/replication_record"}},"created_at":"2026-07-05T11:16:49.072242+00:00","updated_at":"2026-07-05T11:16:49.072242+00:00"}