{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:Z4VOBKGXGFZWTTJPII5LRL5G32","short_pith_number":"pith:Z4VOBKGX","canonical_record":{"source":{"id":"2511.16964","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.MA","submitted_at":"2025-11-21T05:37:38Z","cross_cats_sorted":["cs.AI","cs.DC"],"title_canon_sha256":"08f42973845b6848287e8643acf8abaae8107c35a8a8905d5b8eca6c607c6e44","abstract_canon_sha256":"1ff1dec44e5b433b61e42ec1111f63dac46c3621afeb64856727183cf86e0a1f"},"schema_version":"1.0"},"canonical_sha256":"cf2ae0a8d7317369cd2f423ab8afa6de9bc28f425416eb1a8ce9b3d4dd32297b","source":{"kind":"arxiv","id":"2511.16964","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2511.16964","created_at":"2026-05-17T23:39:17Z"},{"alias_kind":"arxiv_version","alias_value":"2511.16964v2","created_at":"2026-05-17T23:39:17Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.16964","created_at":"2026-05-17T23:39:17Z"},{"alias_kind":"pith_short_12","alias_value":"Z4VOBKGXGFZW","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"Z4VOBKGXGFZWTTJP","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"Z4VOBKGX","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:Z4VOBKGXGFZWTTJPII5LRL5G32","target":"record","payload":{"canonical_record":{"source":{"id":"2511.16964","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.MA","submitted_at":"2025-11-21T05:37:38Z","cross_cats_sorted":["cs.AI","cs.DC"],"title_canon_sha256":"08f42973845b6848287e8643acf8abaae8107c35a8a8905d5b8eca6c607c6e44","abstract_canon_sha256":"1ff1dec44e5b433b61e42ec1111f63dac46c3621afeb64856727183cf86e0a1f"},"schema_version":"1.0"},"canonical_sha256":"cf2ae0a8d7317369cd2f423ab8afa6de9bc28f425416eb1a8ce9b3d4dd32297b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:17.079231Z","signature_b64":"ZG3CqwUEPxtr8GT3Mm7syiftaps8O339GJcY0WCZHcAHU50q9B76cd0Ui3/NtVhpYgTFM6EsqkT10blayn0oBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cf2ae0a8d7317369cd2f423ab8afa6de9bc28f425416eb1a8ce9b3d4dd32297b","last_reissued_at":"2026-05-17T23:39:17.078540Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:17.078540Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2511.16964","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"FumYr7RWVAbCGooopMl5gZ6tmZD2yz3UyYXf8MxO7chSFxDLgiuWcJ6zTEqWoUsDEzj9nrcJF23IZSxuZG5RDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-19T20:47:39.278248Z"},"content_sha256":"d981c7f9eb5191df2eccf4f341a9d62e4ecfadbb9e1796ffab17cbd3a5fd5289","schema_version":"1.0","event_id":"sha256:d981c7f9eb5191df2eccf4f341a9d62e4ecfadbb9e1796ffab17cbd3a5fd5289"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:Z4VOBKGXGFZWTTJPII5LRL5G32","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Optimizing PyTorch Inference with LLM-Based Multi-Agent Systems","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Multi-agent LLM systems optimize PyTorch code for 2.88x faster inference than eager execution on H100 GPUs.","cross_cats":["cs.AI","cs.DC"],"primary_cat":"cs.MA","authors_text":"Costin Iancu, Kirill Nagaitsev, Luka Grbcic, Samuel Williams","submitted_at":"2025-11-21T05:37:38Z","abstract_excerpt":"Maximizing performance on available GPU hardware is an ongoing challenge for modern AI inference systems. Traditional approaches include writing custom GPU kernels and using specialized model compilers to tune high-level code for specific GPU targets. Recent work shows that LLM-based multi-agent systems can effectively perform such tuning, often outperforming existing compilers and eliminating the need for manual kernel development. However, the dynamics of multi-agent systems for this task remain unexplored. In this work, we present a logical framework for comparing multi-agent PyTorch optimi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The best implementation achieves an average 2.88x speedup over PyTorch Eager (1.85x over torch.compile) on an H100 GPU across diverse tasks in KernelBench.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That LLM-based multi-agent systems can reliably generate correct, bug-free optimized code at scale and that results on the KernelBench suite generalize to production inference workloads.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"An exploit-heavy multi-agent LLM system with error-fixing agents delivers 2.88x average speedup over PyTorch Eager and 1.85x over torch.compile on H100 GPUs across KernelBench tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Multi-agent LLM systems optimize PyTorch code for 2.88x faster inference than eager execution on H100 GPUs.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"94eb08b7e71295b5620a6816d68a435978e6ab7a766c505fb1a033f4959806c5"},"source":{"id":"2511.16964","kind":"arxiv","version":2},"verdict":{"id":"ad3137c4-1107-4219-9ed4-ab9604e54428","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T20:59:11.566462Z","strongest_claim":"The best implementation achieves an average 2.88x speedup over PyTorch Eager (1.85x over torch.compile) on an H100 GPU across diverse tasks in KernelBench.","one_line_summary":"An exploit-heavy multi-agent LLM system with error-fixing agents delivers 2.88x average speedup over PyTorch Eager and 1.85x over torch.compile on H100 GPUs across KernelBench tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That LLM-based multi-agent systems can reliably generate correct, bug-free optimized code at scale and that results on the KernelBench suite generalize to production inference workloads.","pith_extraction_headline":"Multi-agent LLM systems optimize PyTorch code for 2.88x faster inference than eager execution on H100 GPUs."},"references":{"count":5,"sample":[{"doi":"10.1145/3703412.3703416","year":2025,"title":"Accessed: 2025-10-17","work_id":"a6173c86-049f-4167-a4fb-a0022b7502c3","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","work_id":"76a0f850-d490-4e4f-ab98-8d25df82cd23","ref_index":2,"cited_arxiv_id":"2506.13131","is_internal_anchor":true},{"doi":"","year":2019,"title":"Anjiang Wei, Tianran Sun, Yogesh Seenichamy, Hang Song, Anne Ouyang, Azalia Mirhoseini, Ke Wang, and Alex Aiken","work_id":"d2dbaa77-84ac-4089-831b-fbbc4af827dc","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Zhang, Z., Bajaj, A","work_id":"03fe7bc1-abe5-47bf-affe-5b84aa98fa39","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"ISBN 9798331314385","work_id":"3e41f374-6f38-4cec-baca-12faade549ba","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":5,"snapshot_sha256":"d4a47fa2f75d1d75d924a93446a58b856d64d12d86e566bcb3b910a69dceba2a","internal_anchors":1},"formal_canon":{"evidence_count":1,"snapshot_sha256":"31063d0985ae57706c90f375287da9de005a80bd14f3b474a89f2e8760c957c1"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"ad3137c4-1107-4219-9ed4-ab9604e54428"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"aSF1M0Ivw2/LmWJlN9h7sunVAff/Wm5NLLuaX7R9fHDLrQr3keE+Sn2sFkrGqhJQveo+sDiqd4w5O4JRTzQlCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-19T20:47:39.278782Z"},"content_sha256":"775048c6b968342a6fe3d74353bab7488008f6d0ff9b9687cabb5faec02234eb","schema_version":"1.0","event_id":"sha256:775048c6b968342a6fe3d74353bab7488008f6d0ff9b9687cabb5faec02234eb"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/Z4VOBKGXGFZWTTJPII5LRL5G32/bundle.json","state_url":"https://pith.science/pith/Z4VOBKGXGFZWTTJPII5LRL5G32/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/Z4VOBKGXGFZWTTJPII5LRL5G32/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-19T20:47:39Z","links":{"resolver":"https://pith.science/pith/Z4VOBKGXGFZWTTJPII5LRL5G32","bundle":"https://pith.science/pith/Z4VOBKGXGFZWTTJPII5LRL5G32/bundle.json","state":"https://pith.science/pith/Z4VOBKGXGFZWTTJPII5LRL5G32/state.json","well_known_bundle":"https://pith.science/.well-known/pith/Z4VOBKGXGFZWTTJPII5LRL5G32/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:Z4VOBKGXGFZWTTJPII5LRL5G32","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"1ff1dec44e5b433b61e42ec1111f63dac46c3621afeb64856727183cf86e0a1f","cross_cats_sorted":["cs.AI","cs.DC"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.MA","submitted_at":"2025-11-21T05:37:38Z","title_canon_sha256":"08f42973845b6848287e8643acf8abaae8107c35a8a8905d5b8eca6c607c6e44"},"schema_version":"1.0","source":{"id":"2511.16964","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2511.16964","created_at":"2026-05-17T23:39:17Z"},{"alias_kind":"arxiv_version","alias_value":"2511.16964v2","created_at":"2026-05-17T23:39:17Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.16964","created_at":"2026-05-17T23:39:17Z"},{"alias_kind":"pith_short_12","alias_value":"Z4VOBKGXGFZW","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"Z4VOBKGXGFZWTTJP","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"Z4VOBKGX","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:775048c6b968342a6fe3d74353bab7488008f6d0ff9b9687cabb5faec02234eb","target":"graph","created_at":"2026-05-17T23:39:17Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"The best implementation achieves an average 2.88x speedup over PyTorch Eager (1.85x over torch.compile) on an H100 GPU across diverse tasks in KernelBench."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That LLM-based multi-agent systems can reliably generate correct, bug-free optimized code at scale and that results on the KernelBench suite generalize to production inference workloads."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"An exploit-heavy multi-agent LLM system with error-fixing agents delivers 2.88x average speedup over PyTorch Eager and 1.85x over torch.compile on H100 GPUs across KernelBench tasks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Multi-agent LLM systems optimize PyTorch code for 2.88x faster inference than eager execution on H100 GPUs."}],"snapshot_sha256":"94eb08b7e71295b5620a6816d68a435978e6ab7a766c505fb1a033f4959806c5"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"31063d0985ae57706c90f375287da9de005a80bd14f3b474a89f2e8760c957c1"},"paper":{"abstract_excerpt":"Maximizing performance on available GPU hardware is an ongoing challenge for modern AI inference systems. Traditional approaches include writing custom GPU kernels and using specialized model compilers to tune high-level code for specific GPU targets. Recent work shows that LLM-based multi-agent systems can effectively perform such tuning, often outperforming existing compilers and eliminating the need for manual kernel development. However, the dynamics of multi-agent systems for this task remain unexplored. In this work, we present a logical framework for comparing multi-agent PyTorch optimi","authors_text":"Costin Iancu, Kirill Nagaitsev, Luka Grbcic, Samuel Williams","cross_cats":["cs.AI","cs.DC"],"headline":"Multi-agent LLM systems optimize PyTorch code for 2.88x faster inference than eager execution on H100 GPUs.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.MA","submitted_at":"2025-11-21T05:37:38Z","title":"Optimizing PyTorch Inference with LLM-Based Multi-Agent Systems"},"references":{"count":5,"internal_anchors":1,"resolved_work":5,"sample":[{"cited_arxiv_id":"","doi":"10.1145/3703412.3703416","is_internal_anchor":false,"ref_index":1,"title":"Accessed: 2025-10-17","work_id":"a6173c86-049f-4167-a4fb-a0022b7502c3","year":2025},{"cited_arxiv_id":"2506.13131","doi":"","is_internal_anchor":true,"ref_index":2,"title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","work_id":"76a0f850-d490-4e4f-ab98-8d25df82cd23","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Anjiang Wei, Tianran Sun, Yogesh Seenichamy, Hang Song, Anne Ouyang, Azalia Mirhoseini, Ke Wang, and Alex Aiken","work_id":"d2dbaa77-84ac-4089-831b-fbbc4af827dc","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Zhang, Z., Bajaj, A","work_id":"03fe7bc1-abe5-47bf-affe-5b84aa98fa39","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"ISBN 9798331314385","work_id":"3e41f374-6f38-4cec-baca-12faade549ba","year":2025}],"snapshot_sha256":"d4a47fa2f75d1d75d924a93446a58b856d64d12d86e566bcb3b910a69dceba2a"},"source":{"id":"2511.16964","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-17T20:59:11.566462Z","id":"ad3137c4-1107-4219-9ed4-ab9604e54428","model_set":{"reader":"grok-4.3"},"one_line_summary":"An exploit-heavy multi-agent LLM system with error-fixing agents delivers 2.88x average speedup over PyTorch Eager and 1.85x over torch.compile on H100 GPUs across KernelBench tasks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Multi-agent LLM systems optimize PyTorch code for 2.88x faster inference than eager execution on H100 GPUs.","strongest_claim":"The best implementation achieves an average 2.88x speedup over PyTorch Eager (1.85x over torch.compile) on an H100 GPU across diverse tasks in KernelBench.","weakest_assumption":"That LLM-based multi-agent systems can reliably generate correct, bug-free optimized code at scale and that results on the KernelBench suite generalize to production inference workloads."}},"verdict_id":"ad3137c4-1107-4219-9ed4-ab9604e54428"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d981c7f9eb5191df2eccf4f341a9d62e4ecfadbb9e1796ffab17cbd3a5fd5289","target":"record","created_at":"2026-05-17T23:39:17Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"1ff1dec44e5b433b61e42ec1111f63dac46c3621afeb64856727183cf86e0a1f","cross_cats_sorted":["cs.AI","cs.DC"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.MA","submitted_at":"2025-11-21T05:37:38Z","title_canon_sha256":"08f42973845b6848287e8643acf8abaae8107c35a8a8905d5b8eca6c607c6e44"},"schema_version":"1.0","source":{"id":"2511.16964","kind":"arxiv","version":2}},"canonical_sha256":"cf2ae0a8d7317369cd2f423ab8afa6de9bc28f425416eb1a8ce9b3d4dd32297b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"cf2ae0a8d7317369cd2f423ab8afa6de9bc28f425416eb1a8ce9b3d4dd32297b","first_computed_at":"2026-05-17T23:39:17.078540Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:17.078540Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZG3CqwUEPxtr8GT3Mm7syiftaps8O339GJcY0WCZHcAHU50q9B76cd0Ui3/NtVhpYgTFM6EsqkT10blayn0oBQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:17.079231Z","signed_message":"canonical_sha256_bytes"},"source_id":"2511.16964","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d981c7f9eb5191df2eccf4f341a9d62e4ecfadbb9e1796ffab17cbd3a5fd5289","sha256:775048c6b968342a6fe3d74353bab7488008f6d0ff9b9687cabb5faec02234eb"],"state_sha256":"43d3442945517b29c5d69fead6f45fcf0ba9550f510f58aef76d4e114f115ede"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kDR57pn1XjgoQqrQcnRWTAE6Um2BOqHOCk3k2a3ucq8lmKjWB2VvNPcymJptENqMcTRvUhyVn+o4Kp2Wt0PPCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-19T20:47:39.281386Z","bundle_sha256":"bce8b77ac5ddbf82d936798a00be1cb75dcdb4a082284e5eff5d4004134740a1"}}