{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:XC46HSFBDQEO3FT2RVX65HMIWN","short_pith_number":"pith:XC46HSFB","canonical_record":{"source":{"id":"2605.07985","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-05-08T16:44:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d776310694207b5c955b74fc08e4fdfe53a7d02b736402c7ec58d1ee6ba53c83","abstract_canon_sha256":"834ef73fc769407acf33c8e36ffc97a63b6e35819da18dbebc1f198915b12784"},"schema_version":"1.0"},"canonical_sha256":"b8b9e3c8a11c08ed967a8d6fee9d88b3720dc03a369c9f993c71a982301ed297","source":{"kind":"arxiv","id":"2605.07985","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.07985","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"arxiv_version","alias_value":"2605.07985v2","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.07985","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"pith_short_12","alias_value":"XC46HSFBDQEO","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"pith_short_16","alias_value":"XC46HSFBDQEO3FT2","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"pith_short_8","alias_value":"XC46HSFB","created_at":"2026-05-22T01:04:05Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:XC46HSFBDQEO3FT2RVX65HMIWN","target":"record","payload":{"canonical_record":{"source":{"id":"2605.07985","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-05-08T16:44:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d776310694207b5c955b74fc08e4fdfe53a7d02b736402c7ec58d1ee6ba53c83","abstract_canon_sha256":"834ef73fc769407acf33c8e36ffc97a63b6e35819da18dbebc1f198915b12784"},"schema_version":"1.0"},"canonical_sha256":"b8b9e3c8a11c08ed967a8d6fee9d88b3720dc03a369c9f993c71a982301ed297","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:05.470461Z","signature_b64":"dj+jIp8lFNZbh8m3jhvnHquVnUNi5vuVa9gdmvfGcIhcvE4bOaYslXjnSNsjhPwq27C4SpP7mA+rO4kbcDxBAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b8b9e3c8a11c08ed967a8d6fee9d88b3720dc03a369c9f993c71a982301ed297","last_reissued_at":"2026-05-22T01:04:05.469584Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:05.469584Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.07985","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":"2026-05-22T01:04:05Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TK4NWiy+vify0ZrGnLHsy45TSIvz8lr11IgxUmwzoGpxC6pWyNA5RWKS4QWaT32cWtrE93z2EeDpUsHmZGpAAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-10T20:20:57.315709Z"},"content_sha256":"f66b476584a02761ada00570ae49e25944363420e9d760cacdcab44e660b2d13","schema_version":"1.0","event_id":"sha256:f66b476584a02761ada00570ae49e25944363420e9d760cacdcab44e660b2d13"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:XC46HSFBDQEO3FT2RVX65HMIWN","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Dooly: Configuration-Agnostic, Redundancy-Aware Profiling for LLM Inference Simulation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Dooly reuses profiled LLM operation latencies across model configurations by tracing input-dimension origins in a single inference pass.","cross_cats":["cs.AI"],"primary_cat":"cs.DC","authors_text":"Anoop Rachakonda, Daehyeok Kim, Geon-Woo Kim, Joon Ha Kim","submitted_at":"2026-05-08T16:44:47Z","abstract_excerpt":"Selecting the optimal LLM inference configuration requires evaluation across hardware, serving engines, attention backends, and model architectures, since no single choice performs best across all workloads. Profile-based simulators are the standard tool, yet they hardcode their operation set to a specific configuration and re-profile every operation from scratch, making exploration prohibitively expensive. This cost stems from a missing structural understanding: every input dimension of each operation is fixed by the model configuration or determined by the incoming request. Many model-config"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across two GPU platforms, three attention backends, and diverse model architectures, Dooly achieves simulation accuracy within 5% MAPE for TTFT and 8% for TPOT while reducing profiling GPU-hours by 56.4% across 12 models compared to the existing profiling approach.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That labeling input dimensions by origin via taint propagation during one inference pass is sufficient to identify all reusable operations across arbitrary model configurations, and that the resulting latency database plus regression models generalize without hidden dependencies or post-hoc tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Dooly reduces LLM inference profiling costs by 56.4% via configuration-agnostic taint-based labeling and selective database reuse, delivering simulation accuracy within 5% MAPE for TTFT and 8% for TPOT across 12 models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Dooly reuses profiled LLM operation latencies across model configurations by tracing input-dimension origins in a single inference pass.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"043f8e767920a65291bcbdc2dbc725dccda0e5d716ce79470f199ca7980f1c10"},"source":{"id":"2605.07985","kind":"arxiv","version":2},"verdict":{"id":"1a80f643-4890-4af7-8680-5fb670039996","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T02:37:15.956351Z","strongest_claim":"Across two GPU platforms, three attention backends, and diverse model architectures, Dooly achieves simulation accuracy within 5% MAPE for TTFT and 8% for TPOT while reducing profiling GPU-hours by 56.4% across 12 models compared to the existing profiling approach.","one_line_summary":"Dooly reduces LLM inference profiling costs by 56.4% via configuration-agnostic taint-based labeling and selective database reuse, delivering simulation accuracy within 5% MAPE for TTFT and 8% for TPOT across 12 models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That labeling input dimensions by origin via taint propagation during one inference pass is sufficient to identify all reusable operations across arbitrary model configurations, and that the resulting latency database plus regression models generalize without hidden dependencies or post-hoc tuning.","pith_extraction_headline":"Dooly reuses profiled LLM operation latencies across model configurations by tracing input-dimension origins in a single inference pass."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.07985/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T10:02:08.726065Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-20T04:44:29.175780Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T15:31:17.998249Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T11:20:50.210815Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"304272161f2f5db4e9a56f93633e2f7b2a40886da088de7224e24b930f6e15bc"},"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"},"verdict_id":"1a80f643-4890-4af7-8680-5fb670039996"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-22T01:04:05Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"sATXHhe2LgsQDmLSRF4MRLAFRrIEMg4HJHTIdyhscroQlP6N22OUYc2LmGq+P7d5BEpC1I0lmaAEblVCtt9+BA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-10T20:20:57.316208Z"},"content_sha256":"e717964b734ffbed7e834f5736e8d3d70b6b4bd291a30233ce91c70f347005c5","schema_version":"1.0","event_id":"sha256:e717964b734ffbed7e834f5736e8d3d70b6b4bd291a30233ce91c70f347005c5"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/XC46HSFBDQEO3FT2RVX65HMIWN/bundle.json","state_url":"https://pith.science/pith/XC46HSFBDQEO3FT2RVX65HMIWN/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/XC46HSFBDQEO3FT2RVX65HMIWN/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":"2026-07-10T20:20:57Z","links":{"resolver":"https://pith.science/pith/XC46HSFBDQEO3FT2RVX65HMIWN","bundle":"https://pith.science/pith/XC46HSFBDQEO3FT2RVX65HMIWN/bundle.json","state":"https://pith.science/pith/XC46HSFBDQEO3FT2RVX65HMIWN/state.json","well_known_bundle":"https://pith.science/.well-known/pith/XC46HSFBDQEO3FT2RVX65HMIWN/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:XC46HSFBDQEO3FT2RVX65HMIWN","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":"834ef73fc769407acf33c8e36ffc97a63b6e35819da18dbebc1f198915b12784","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-05-08T16:44:47Z","title_canon_sha256":"d776310694207b5c955b74fc08e4fdfe53a7d02b736402c7ec58d1ee6ba53c83"},"schema_version":"1.0","source":{"id":"2605.07985","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.07985","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"arxiv_version","alias_value":"2605.07985v2","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.07985","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"pith_short_12","alias_value":"XC46HSFBDQEO","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"pith_short_16","alias_value":"XC46HSFBDQEO3FT2","created_at":"2026-05-22T01:04:05Z"},{"alias_kind":"pith_short_8","alias_value":"XC46HSFB","created_at":"2026-05-22T01:04:05Z"}],"graph_snapshots":[{"event_id":"sha256:e717964b734ffbed7e834f5736e8d3d70b6b4bd291a30233ce91c70f347005c5","target":"graph","created_at":"2026-05-22T01:04:05Z","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":"Across two GPU platforms, three attention backends, and diverse model architectures, Dooly achieves simulation accuracy within 5% MAPE for TTFT and 8% for TPOT while reducing profiling GPU-hours by 56.4% across 12 models compared to the existing profiling approach."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That labeling input dimensions by origin via taint propagation during one inference pass is sufficient to identify all reusable operations across arbitrary model configurations, and that the resulting latency database plus regression models generalize without hidden dependencies or post-hoc tuning."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Dooly reduces LLM inference profiling costs by 56.4% via configuration-agnostic taint-based labeling and selective database reuse, delivering simulation accuracy within 5% MAPE for TTFT and 8% for TPOT across 12 models."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Dooly reuses profiled LLM operation latencies across model configurations by tracing input-dimension origins in a single inference pass."}],"snapshot_sha256":"043f8e767920a65291bcbdc2dbc725dccda0e5d716ce79470f199ca7980f1c10"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-20T10:02:08.726065Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T04:44:29.175780Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T15:31:17.998249Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T11:20:50.210815Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.07985/integrity.json","findings":[],"snapshot_sha256":"304272161f2f5db4e9a56f93633e2f7b2a40886da088de7224e24b930f6e15bc","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Selecting the optimal LLM inference configuration requires evaluation across hardware, serving engines, attention backends, and model architectures, since no single choice performs best across all workloads. Profile-based simulators are the standard tool, yet they hardcode their operation set to a specific configuration and re-profile every operation from scratch, making exploration prohibitively expensive. This cost stems from a missing structural understanding: every input dimension of each operation is fixed by the model configuration or determined by the incoming request. Many model-config","authors_text":"Anoop Rachakonda, Daehyeok Kim, Geon-Woo Kim, Joon Ha Kim","cross_cats":["cs.AI"],"headline":"Dooly reuses profiled LLM operation latencies across model configurations by tracing input-dimension origins in a single inference pass.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-05-08T16:44:47Z","title":"Dooly: Configuration-Agnostic, Redundancy-Aware Profiling for LLM Inference Simulation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.07985","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-11T02:37:15.956351Z","id":"1a80f643-4890-4af7-8680-5fb670039996","model_set":{"reader":"grok-4.3"},"one_line_summary":"Dooly reduces LLM inference profiling costs by 56.4% via configuration-agnostic taint-based labeling and selective database reuse, delivering simulation accuracy within 5% MAPE for TTFT and 8% for TPOT across 12 models.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Dooly reuses profiled LLM operation latencies across model configurations by tracing input-dimension origins in a single inference pass.","strongest_claim":"Across two GPU platforms, three attention backends, and diverse model architectures, Dooly achieves simulation accuracy within 5% MAPE for TTFT and 8% for TPOT while reducing profiling GPU-hours by 56.4% across 12 models compared to the existing profiling approach.","weakest_assumption":"That labeling input dimensions by origin via taint propagation during one inference pass is sufficient to identify all reusable operations across arbitrary model configurations, and that the resulting latency database plus regression models generalize without hidden dependencies or post-hoc tuning."}},"verdict_id":"1a80f643-4890-4af7-8680-5fb670039996"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f66b476584a02761ada00570ae49e25944363420e9d760cacdcab44e660b2d13","target":"record","created_at":"2026-05-22T01:04:05Z","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":"834ef73fc769407acf33c8e36ffc97a63b6e35819da18dbebc1f198915b12784","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-05-08T16:44:47Z","title_canon_sha256":"d776310694207b5c955b74fc08e4fdfe53a7d02b736402c7ec58d1ee6ba53c83"},"schema_version":"1.0","source":{"id":"2605.07985","kind":"arxiv","version":2}},"canonical_sha256":"b8b9e3c8a11c08ed967a8d6fee9d88b3720dc03a369c9f993c71a982301ed297","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b8b9e3c8a11c08ed967a8d6fee9d88b3720dc03a369c9f993c71a982301ed297","first_computed_at":"2026-05-22T01:04:05.469584Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-22T01:04:05.469584Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"dj+jIp8lFNZbh8m3jhvnHquVnUNi5vuVa9gdmvfGcIhcvE4bOaYslXjnSNsjhPwq27C4SpP7mA+rO4kbcDxBAg==","signature_status":"signed_v1","signed_at":"2026-05-22T01:04:05.470461Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.07985","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f66b476584a02761ada00570ae49e25944363420e9d760cacdcab44e660b2d13","sha256:e717964b734ffbed7e834f5736e8d3d70b6b4bd291a30233ce91c70f347005c5"],"state_sha256":"9faa4b80fd4236163254cdb2703b900eda19724e5cea2501efe0121d50178d28"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"n2xinaj4F17kf6JCAzSRJmRk7zyErM44BZNJBWzFs2PO1Pxzc739XtVAGJYTtCv31fsAnOIqn5XJc49Z/E+nBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-10T20:20:57.318822Z","bundle_sha256":"a7362eff5ee99e49b8cb44d52532c35b4e6ab1e4b6468ca6c96f9312341d8204"}}