{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DKGHEJAXS43JIYWIZESD4EEDJ5","short_pith_number":"pith:DKGHEJAX","schema_version":"1.0","canonical_sha256":"1a8c72241797369462c8c9243e10834f5210b8509ee0df4a25feb17ab71f7696","source":{"kind":"arxiv","id":"2605.31464","version":1},"attestation_state":"computed","paper":{"title":"GPU Forecasters: Language Models as Selective Surrogates for Kernel Runtime Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Elias Stengel-Eskin, Jaemin Cho, Justin Chih-Yao Chen, Mohit Bansal, Zaid Khan","submitted_at":"2026-05-29T15:56:08Z","abstract_excerpt":"GPU kernels are the workhorse of modern deep learning, and optimizing them (via evolutionary search or coding agents) usually requires repeated measurement on target hardware. While these measurements provide the ground-truth signal necessary for kernel search, they are costly, because each evaluation of a kernel requires compilation and repeated execution on a GPU. As improvements in LLM inference reduce the cost of writing novel kernels and LLM-driven searches scale to large search budgets, on-device evaluation becomes a bottleneck. To address this, we study how LLMs can serve as selective G"},"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":"2605.31464","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-29T15:56:08Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ab92b8d73712121e3f669cbc56ed51c99abdadd619411e80b12046d2c9ae914b","abstract_canon_sha256":"b5a00d4efa03ecbcdd4026e9fcea01f955ddac7f33ea3cd910384a62b3dcd271"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T02:04:07.365901Z","signature_b64":"5Sru9DBjVuCgAQXzlYTa+MkO/Mv5yNAFzi/cS3FjEiDQz1UIs+BJG8f2jRAO21NBl/2GHiS79sEI6NG2q7eAAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1a8c72241797369462c8c9243e10834f5210b8509ee0df4a25feb17ab71f7696","last_reissued_at":"2026-06-01T02:04:07.365107Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T02:04:07.365107Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GPU Forecasters: Language Models as Selective Surrogates for Kernel Runtime Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Elias Stengel-Eskin, Jaemin Cho, Justin Chih-Yao Chen, Mohit Bansal, Zaid Khan","submitted_at":"2026-05-29T15:56:08Z","abstract_excerpt":"GPU kernels are the workhorse of modern deep learning, and optimizing them (via evolutionary search or coding agents) usually requires repeated measurement on target hardware. While these measurements provide the ground-truth signal necessary for kernel search, they are costly, because each evaluation of a kernel requires compilation and repeated execution on a GPU. As improvements in LLM inference reduce the cost of writing novel kernels and LLM-driven searches scale to large search budgets, on-device evaluation becomes a bottleneck. To address this, we study how LLMs can serve as selective G"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.31464","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/2605.31464/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":"2605.31464","created_at":"2026-06-01T02:04:07.365232+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.31464v1","created_at":"2026-06-01T02:04:07.365232+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.31464","created_at":"2026-06-01T02:04:07.365232+00:00"},{"alias_kind":"pith_short_12","alias_value":"DKGHEJAXS43J","created_at":"2026-06-01T02:04:07.365232+00:00"},{"alias_kind":"pith_short_16","alias_value":"DKGHEJAXS43JIYWI","created_at":"2026-06-01T02:04:07.365232+00:00"},{"alias_kind":"pith_short_8","alias_value":"DKGHEJAX","created_at":"2026-06-01T02:04:07.365232+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5","json":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5.json","graph_json":"https://pith.science/api/pith-number/DKGHEJAXS43JIYWIZESD4EEDJ5/graph.json","events_json":"https://pith.science/api/pith-number/DKGHEJAXS43JIYWIZESD4EEDJ5/events.json","paper":"https://pith.science/paper/DKGHEJAX"},"agent_actions":{"view_html":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5","download_json":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5.json","view_paper":"https://pith.science/paper/DKGHEJAX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.31464&json=true","fetch_graph":"https://pith.science/api/pith-number/DKGHEJAXS43JIYWIZESD4EEDJ5/graph.json","fetch_events":"https://pith.science/api/pith-number/DKGHEJAXS43JIYWIZESD4EEDJ5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5/action/storage_attestation","attest_author":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5/action/author_attestation","sign_citation":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5/action/citation_signature","submit_replication":"https://pith.science/pith/DKGHEJAXS43JIYWIZESD4EEDJ5/action/replication_record"}},"created_at":"2026-06-01T02:04:07.365232+00:00","updated_at":"2026-06-01T02:04:07.365232+00:00"}