{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:FO7JZTPBJLUMS37IKFOZVT3GGA","short_pith_number":"pith:FO7JZTPB","canonical_record":{"source":{"id":"2603.15757","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-03-16T18:00:19Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"413ec9f41f749213d22548e7fbbe130d74531308fb7dd4415c79cf0d44cb2d63","abstract_canon_sha256":"9066531862b8bf1340e300a9dba0bb55df9f8c9fbcf50cdc74e16747831d43f1"},"schema_version":"1.0"},"canonical_sha256":"2bbe9ccde14ae8c96fe8515d9acf66303ee30d42fd4d73c076ab5f62ef67987a","source":{"kind":"arxiv","id":"2603.15757","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.15757","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"arxiv_version","alias_value":"2603.15757v3","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.15757","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"pith_short_12","alias_value":"FO7JZTPBJLUM","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"pith_short_16","alias_value":"FO7JZTPBJLUMS37I","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"pith_short_8","alias_value":"FO7JZTPB","created_at":"2026-06-23T01:13:04Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:FO7JZTPBJLUMS37IKFOZVT3GGA","target":"record","payload":{"canonical_record":{"source":{"id":"2603.15757","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-03-16T18:00:19Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"413ec9f41f749213d22548e7fbbe130d74531308fb7dd4415c79cf0d44cb2d63","abstract_canon_sha256":"9066531862b8bf1340e300a9dba0bb55df9f8c9fbcf50cdc74e16747831d43f1"},"schema_version":"1.0"},"canonical_sha256":"2bbe9ccde14ae8c96fe8515d9acf66303ee30d42fd4d73c076ab5f62ef67987a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T01:13:04.365075Z","signature_b64":"ZKOcuopqR2+YDR/6Rul0WHIRuEKbMZ5+uYV9YD7SgY8jRMQnK+SGwPyF4MxPPJSnZ1TamUzmu+N10+Slgi7VCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2bbe9ccde14ae8c96fe8515d9acf66303ee30d42fd4d73c076ab5f62ef67987a","last_reissued_at":"2026-06-23T01:13:04.364579Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T01:13:04.364579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2603.15757","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-23T01:13:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"R0v5yXjaptPysC1l1L+F81eymcvgY5Hf6Dr+KcAph5uCDjZya2ANForiQOlLpI9Z7Mh9h/JvDedIyJWWMLDtCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T07:00:23.724791Z"},"content_sha256":"d434025fbd6e709e1c72195ed5f19302109f5dcf40ba727c40cf6c70c526e19b","schema_version":"1.0","event_id":"sha256:d434025fbd6e709e1c72195ed5f19302109f5dcf40ba727c40cf6c70c526e19b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:FO7JZTPBJLUMS37IKFOZVT3GGA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"You've Got a Golden Ticket: Improving Generative Robot Policies With A Single Noise Vector","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A fixed initial noise vector can improve the performance of pretrained diffusion and flow-matching robot policies without any retraining.","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Eric Rosen, Karl Schmeckpeper, Kausik Sivakumar, Nakul Gopalan, Omkar Patil, Ondrej Biza, Robin Walters, Sebastian Castro, Stephen Hart, Thomas Weng, Wil Thomason, Xiaohan Zhang","submitted_at":"2026-03-16T18:00:19Z","abstract_excerpt":"What happens when a pretrained generative robot policy is provided a constant initial noise as input, rather than repeatedly sampling it from a Gaussian? We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input - a golden ticket. We propose simple search methods to find golden tickets using Monte-Carlo policy evaluation that keeps the pretrained policy frozen"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input -- a golden ticket.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a noise vector found by Monte-Carlo search on a modest number of rollouts will continue to produce higher reward on new episodes and under distribution shift without the policy being updated or the search overfitting to the evaluation episodes used during search.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Optimizing a single constant initial noise vector for frozen generative robot policies improves success rates on 38 of 43 tasks by up to 58% relative improvement.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A fixed initial noise vector can improve the performance of pretrained diffusion and flow-matching robot policies without any retraining.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"13af59707b3505a2888ff1b154aceabe072358a0c653ba50988fce9410ad2797"},"source":{"id":"2603.15757","kind":"arxiv","version":3},"verdict":{"id":"7091a265-ba31-41a3-b887-73f1715b13ab","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T09:50:08.193482Z","strongest_claim":"We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input -- a golden ticket.","one_line_summary":"Optimizing a single constant initial noise vector for frozen generative robot policies improves success rates on 38 of 43 tasks by up to 58% relative improvement.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a noise vector found by Monte-Carlo search on a modest number of rollouts will continue to produce higher reward on new episodes and under distribution shift without the policy being updated or the search overfitting to the evaluation episodes used during search.","pith_extraction_headline":"A fixed initial noise vector can improve the performance of pretrained diffusion and flow-matching robot policies without any retraining."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.15757/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":1,"snapshot_sha256":"63f8f57c545556ee0acee5eabfa7c7f973cc006518505095252593a57459330c"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"7091a265-ba31-41a3-b887-73f1715b13ab"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-23T01:13:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kKIzCbLUjOApiPU7zSjWQoQMopKnBtMgmUUm7g7WwXWQIzEr+sZT/P3MVEsNFSHikVZNA/erlMLFUlf5DDOtBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T07:00:23.725520Z"},"content_sha256":"68b2278babb8338507f555199da7e998e536a27f21f04481a1244af208482319","schema_version":"1.0","event_id":"sha256:68b2278babb8338507f555199da7e998e536a27f21f04481a1244af208482319"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FO7JZTPBJLUMS37IKFOZVT3GGA/bundle.json","state_url":"https://pith.science/pith/FO7JZTPBJLUMS37IKFOZVT3GGA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FO7JZTPBJLUMS37IKFOZVT3GGA/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":"2026-08-05T07:00:23Z","links":{"resolver":"https://pith.science/pith/FO7JZTPBJLUMS37IKFOZVT3GGA","bundle":"https://pith.science/pith/FO7JZTPBJLUMS37IKFOZVT3GGA/bundle.json","state":"https://pith.science/pith/FO7JZTPBJLUMS37IKFOZVT3GGA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FO7JZTPBJLUMS37IKFOZVT3GGA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:FO7JZTPBJLUMS37IKFOZVT3GGA","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":"9066531862b8bf1340e300a9dba0bb55df9f8c9fbcf50cdc74e16747831d43f1","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-03-16T18:00:19Z","title_canon_sha256":"413ec9f41f749213d22548e7fbbe130d74531308fb7dd4415c79cf0d44cb2d63"},"schema_version":"1.0","source":{"id":"2603.15757","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.15757","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"arxiv_version","alias_value":"2603.15757v3","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.15757","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"pith_short_12","alias_value":"FO7JZTPBJLUM","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"pith_short_16","alias_value":"FO7JZTPBJLUMS37I","created_at":"2026-06-23T01:13:04Z"},{"alias_kind":"pith_short_8","alias_value":"FO7JZTPB","created_at":"2026-06-23T01:13:04Z"}],"graph_snapshots":[{"event_id":"sha256:68b2278babb8338507f555199da7e998e536a27f21f04481a1244af208482319","target":"graph","created_at":"2026-06-23T01:13:04Z","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":"We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input -- a golden ticket."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That a noise vector found by Monte-Carlo search on a modest number of rollouts will continue to produce higher reward on new episodes and under distribution shift without the policy being updated or the search overfitting to the evaluation episodes used during search."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Optimizing a single constant initial noise vector for frozen generative robot policies improves success rates on 38 of 43 tasks by up to 58% relative improvement."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A fixed initial noise vector can improve the performance of pretrained diffusion and flow-matching robot policies without any retraining."}],"snapshot_sha256":"13af59707b3505a2888ff1b154aceabe072358a0c653ba50988fce9410ad2797"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"63f8f57c545556ee0acee5eabfa7c7f973cc006518505095252593a57459330c"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2603.15757/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"What happens when a pretrained generative robot policy is provided a constant initial noise as input, rather than repeatedly sampling it from a Gaussian? We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input - a golden ticket. We propose simple search methods to find golden tickets using Monte-Carlo policy evaluation that keeps the pretrained policy frozen","authors_text":"Eric Rosen, Karl Schmeckpeper, Kausik Sivakumar, Nakul Gopalan, Omkar Patil, Ondrej Biza, Robin Walters, Sebastian Castro, Stephen Hart, Thomas Weng, Wil Thomason, Xiaohan Zhang","cross_cats":["cs.AI"],"headline":"A fixed initial noise vector can improve the performance of pretrained diffusion and flow-matching robot policies without any retraining.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-03-16T18:00:19Z","title":"You've Got a Golden Ticket: Improving Generative Robot Policies With A Single Noise Vector"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.15757","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-15T09:50:08.193482Z","id":"7091a265-ba31-41a3-b887-73f1715b13ab","model_set":{"reader":"grok-4.3"},"one_line_summary":"Optimizing a single constant initial noise vector for frozen generative robot policies improves success rates on 38 of 43 tasks by up to 58% relative improvement.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A fixed initial noise vector can improve the performance of pretrained diffusion and flow-matching robot policies without any retraining.","strongest_claim":"We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input -- a golden ticket.","weakest_assumption":"That a noise vector found by Monte-Carlo search on a modest number of rollouts will continue to produce higher reward on new episodes and under distribution shift without the policy being updated or the search overfitting to the evaluation episodes used during search."}},"verdict_id":"7091a265-ba31-41a3-b887-73f1715b13ab"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d434025fbd6e709e1c72195ed5f19302109f5dcf40ba727c40cf6c70c526e19b","target":"record","created_at":"2026-06-23T01:13:04Z","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":"9066531862b8bf1340e300a9dba0bb55df9f8c9fbcf50cdc74e16747831d43f1","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-03-16T18:00:19Z","title_canon_sha256":"413ec9f41f749213d22548e7fbbe130d74531308fb7dd4415c79cf0d44cb2d63"},"schema_version":"1.0","source":{"id":"2603.15757","kind":"arxiv","version":3}},"canonical_sha256":"2bbe9ccde14ae8c96fe8515d9acf66303ee30d42fd4d73c076ab5f62ef67987a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2bbe9ccde14ae8c96fe8515d9acf66303ee30d42fd4d73c076ab5f62ef67987a","first_computed_at":"2026-06-23T01:13:04.364579Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-23T01:13:04.364579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZKOcuopqR2+YDR/6Rul0WHIRuEKbMZ5+uYV9YD7SgY8jRMQnK+SGwPyF4MxPPJSnZ1TamUzmu+N10+Slgi7VCA==","signature_status":"signed_v1","signed_at":"2026-06-23T01:13:04.365075Z","signed_message":"canonical_sha256_bytes"},"source_id":"2603.15757","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d434025fbd6e709e1c72195ed5f19302109f5dcf40ba727c40cf6c70c526e19b","sha256:68b2278babb8338507f555199da7e998e536a27f21f04481a1244af208482319"],"state_sha256":"879a6fdff80d84b7c0a961dbbee7bcc038fb066dd178708404e2440a440770e8"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XW1e2L/sUK4ArIathEpsK2zfk1lso6CUUVPsbIkkevNa5Db4XCagwNTYOGWAhxCK21WnO2BjLlO/G/kwQL4tAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T07:00:23.730895Z","bundle_sha256":"22daaf45ccbe9b7bdc3582507277b67df8e4e16b8535083f9a31cd1240ab9e26"}}