{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:C2RCSGB7M6EQ76I6YSX7T3PFCL","short_pith_number":"pith:C2RCSGB7","canonical_record":{"source":{"id":"2602.05993","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-05T18:42:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dce64a21aac61e231e8e36eab9d8f603626df5934320c2350d1387d7f151ed80","abstract_canon_sha256":"e582ab735b562c3b963a8c8e9c6f304e7968beacd9e8d48d40b10731e97eae94"},"schema_version":"1.0"},"canonical_sha256":"16a229183f67890ff91ec4aff9ede512cdec4dca8109d454c797a176a3798b9b","source":{"kind":"arxiv","id":"2602.05993","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.05993","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"arxiv_version","alias_value":"2602.05993v3","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.05993","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"pith_short_12","alias_value":"C2RCSGB7M6EQ","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"pith_short_16","alias_value":"C2RCSGB7M6EQ76I6","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"pith_short_8","alias_value":"C2RCSGB7","created_at":"2026-05-20T00:03:04Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:C2RCSGB7M6EQ76I6YSX7T3PFCL","target":"record","payload":{"canonical_record":{"source":{"id":"2602.05993","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-05T18:42:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dce64a21aac61e231e8e36eab9d8f603626df5934320c2350d1387d7f151ed80","abstract_canon_sha256":"e582ab735b562c3b963a8c8e9c6f304e7968beacd9e8d48d40b10731e97eae94"},"schema_version":"1.0"},"canonical_sha256":"16a229183f67890ff91ec4aff9ede512cdec4dca8109d454c797a176a3798b9b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:04.991138Z","signature_b64":"ZCyOFwFXuASkKir50diL+M7kjX6PwvvIbw5zAyz0frwDO9zAi6v5nZdq32JB+iHdcW6JtnAmNrHPXha3up/nCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"16a229183f67890ff91ec4aff9ede512cdec4dca8109d454c797a176a3798b9b","last_reissued_at":"2026-05-20T00:03:04.990361Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:04.990361Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2602.05993","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":"2026-05-20T00:03:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WvXs3+TWE7fNFSvDBLPObeqcuTDk/JCyplBi2voOPWx+spo9ff7+oVEwzJZVBqqbiOUkkXoT1olioHm4g8VvCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-10T12:27:18.523248Z"},"content_sha256":"2902921a8d0cf84ce2ff1f0dd56ceb1eb8b4f663e04a5685be2441dd502281cc","schema_version":"1.0","event_id":"sha256:2902921a8d0cf84ce2ff1f0dd56ceb1eb8b4f663e04a5685be2441dd502281cc"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:C2RCSGB7M6EQ76I6YSX7T3PFCL","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Diamond Maps: Efficient Reward Alignment via Stochastic Flow Maps","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Diamond Maps turn multi-step stochastic flows into single-step samplers that still support optimal reward alignment at inference time.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Douglas Chen, Giri Anantharaman, Ishin Shah, Luca Eyring, Max Simchowitz, Nicholas Matthew Boffi, Peter Holderrieth, Tommi Jaakkola, Yutong He, Zeynep Akata","submitted_at":"2026-02-05T18:42:00Z","abstract_excerpt":"Flow and diffusion models produce high-quality samples, but adapting them to user preferences or constraints post-training remains costly and brittle, a challenge commonly called reward alignment. We argue that efficient reward alignment should be a property of the generative model itself, not an afterthought, and redesign the model for adaptability. We propose \"Diamond Maps\", stochastic flow map models that enable efficient and accurate alignment to arbitrary rewards at inference time. Diamond Maps amortize many simulation steps into a single-step sampler, like flow maps, while preserving the"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Diamond Maps amortize many simulation steps into a single-step sampler while preserving the stochasticity required for optimal reward alignment, enabling efficient and consistent estimation of the value function for search, SMC, and guidance.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That distilling from GLASS Flows into a stochastic single-step map preserves enough of the original multi-step dynamics to support optimal reward alignment without introducing bias or losing expressivity in the value function estimates.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Diamond Maps are stochastic flow maps that enable efficient, accurate reward alignment for flow and diffusion models at inference time via distillation from GLASS Flows.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Diamond Maps turn multi-step stochastic flows into single-step samplers that still support optimal reward alignment at inference time.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c8c3512a1043d67ed323f967568a92df25596ccc80df2a88c10a1f6ce3f21e45"},"source":{"id":"2602.05993","kind":"arxiv","version":3},"verdict":{"id":"646c8caf-128a-46f9-a7b7-d7fce635e9da","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T06:43:23.627957Z","strongest_claim":"Diamond Maps amortize many simulation steps into a single-step sampler while preserving the stochasticity required for optimal reward alignment, enabling efficient and consistent estimation of the value function for search, SMC, and guidance.","one_line_summary":"Diamond Maps are stochastic flow maps that enable efficient, accurate reward alignment for flow and diffusion models at inference time via distillation from GLASS Flows.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That distilling from GLASS Flows into a stochastic single-step map preserves enough of the original multi-step dynamics to support optimal reward alignment without introducing bias or losing expressivity in the value function estimates.","pith_extraction_headline":"Diamond Maps turn multi-step stochastic flows into single-step samplers that still support optimal reward alignment at inference time."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.05993/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":2,"snapshot_sha256":"d4f355e8996c4d898edbf8645243085610f8c3951d3d353b4159241a926e023f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"646c8caf-128a-46f9-a7b7-d7fce635e9da"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"G94W3IHOoSJkA0sxu/b/7w9Rq8zQXj4fSAolXQdloCfPaMq3RLfVWB4cpqzIuyoAlOHULrEuITdmktCmBJ8cDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-10T12:27:18.524225Z"},"content_sha256":"0b307cc08be91bc215dfad16e169a16c794359ecb00b469a19ddd446661b573b","schema_version":"1.0","event_id":"sha256:0b307cc08be91bc215dfad16e169a16c794359ecb00b469a19ddd446661b573b"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL/bundle.json","state_url":"https://pith.science/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL/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":"2026-06-10T12:27:18Z","links":{"resolver":"https://pith.science/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL","bundle":"https://pith.science/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL/bundle.json","state":"https://pith.science/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL/state.json","well_known_bundle":"https://pith.science/.well-known/pith/C2RCSGB7M6EQ76I6YSX7T3PFCL/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:C2RCSGB7M6EQ76I6YSX7T3PFCL","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":"e582ab735b562c3b963a8c8e9c6f304e7968beacd9e8d48d40b10731e97eae94","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-05T18:42:00Z","title_canon_sha256":"dce64a21aac61e231e8e36eab9d8f603626df5934320c2350d1387d7f151ed80"},"schema_version":"1.0","source":{"id":"2602.05993","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.05993","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"arxiv_version","alias_value":"2602.05993v3","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.05993","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"pith_short_12","alias_value":"C2RCSGB7M6EQ","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"pith_short_16","alias_value":"C2RCSGB7M6EQ76I6","created_at":"2026-05-20T00:03:04Z"},{"alias_kind":"pith_short_8","alias_value":"C2RCSGB7","created_at":"2026-05-20T00:03:04Z"}],"graph_snapshots":[{"event_id":"sha256:0b307cc08be91bc215dfad16e169a16c794359ecb00b469a19ddd446661b573b","target":"graph","created_at":"2026-05-20T00:03:04Z","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":"Diamond Maps amortize many simulation steps into a single-step sampler while preserving the stochasticity required for optimal reward alignment, enabling efficient and consistent estimation of the value function for search, SMC, and guidance."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That distilling from GLASS Flows into a stochastic single-step map preserves enough of the original multi-step dynamics to support optimal reward alignment without introducing bias or losing expressivity in the value function estimates."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Diamond Maps are stochastic flow maps that enable efficient, accurate reward alignment for flow and diffusion models at inference time via distillation from GLASS Flows."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Diamond Maps turn multi-step stochastic flows into single-step samplers that still support optimal reward alignment at inference time."}],"snapshot_sha256":"c8c3512a1043d67ed323f967568a92df25596ccc80df2a88c10a1f6ce3f21e45"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"d4f355e8996c4d898edbf8645243085610f8c3951d3d353b4159241a926e023f"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2602.05993/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Flow and diffusion models produce high-quality samples, but adapting them to user preferences or constraints post-training remains costly and brittle, a challenge commonly called reward alignment. We argue that efficient reward alignment should be a property of the generative model itself, not an afterthought, and redesign the model for adaptability. We propose \"Diamond Maps\", stochastic flow map models that enable efficient and accurate alignment to arbitrary rewards at inference time. Diamond Maps amortize many simulation steps into a single-step sampler, like flow maps, while preserving the","authors_text":"Douglas Chen, Giri Anantharaman, Ishin Shah, Luca Eyring, Max Simchowitz, Nicholas Matthew Boffi, Peter Holderrieth, Tommi Jaakkola, Yutong He, Zeynep Akata","cross_cats":["cs.AI"],"headline":"Diamond Maps turn multi-step stochastic flows into single-step samplers that still support optimal reward alignment at inference time.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-05T18:42:00Z","title":"Diamond Maps: Efficient Reward Alignment via Stochastic Flow Maps"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.05993","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-16T06:43:23.627957Z","id":"646c8caf-128a-46f9-a7b7-d7fce635e9da","model_set":{"reader":"grok-4.3"},"one_line_summary":"Diamond Maps are stochastic flow maps that enable efficient, accurate reward alignment for flow and diffusion models at inference time via distillation from GLASS Flows.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Diamond Maps turn multi-step stochastic flows into single-step samplers that still support optimal reward alignment at inference time.","strongest_claim":"Diamond Maps amortize many simulation steps into a single-step sampler while preserving the stochasticity required for optimal reward alignment, enabling efficient and consistent estimation of the value function for search, SMC, and guidance.","weakest_assumption":"That distilling from GLASS Flows into a stochastic single-step map preserves enough of the original multi-step dynamics to support optimal reward alignment without introducing bias or losing expressivity in the value function estimates."}},"verdict_id":"646c8caf-128a-46f9-a7b7-d7fce635e9da"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:2902921a8d0cf84ce2ff1f0dd56ceb1eb8b4f663e04a5685be2441dd502281cc","target":"record","created_at":"2026-05-20T00:03:04Z","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":"e582ab735b562c3b963a8c8e9c6f304e7968beacd9e8d48d40b10731e97eae94","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-05T18:42:00Z","title_canon_sha256":"dce64a21aac61e231e8e36eab9d8f603626df5934320c2350d1387d7f151ed80"},"schema_version":"1.0","source":{"id":"2602.05993","kind":"arxiv","version":3}},"canonical_sha256":"16a229183f67890ff91ec4aff9ede512cdec4dca8109d454c797a176a3798b9b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"16a229183f67890ff91ec4aff9ede512cdec4dca8109d454c797a176a3798b9b","first_computed_at":"2026-05-20T00:03:04.990361Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:04.990361Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZCyOFwFXuASkKir50diL+M7kjX6PwvvIbw5zAyz0frwDO9zAi6v5nZdq32JB+iHdcW6JtnAmNrHPXha3up/nCg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:04.991138Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.05993","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2902921a8d0cf84ce2ff1f0dd56ceb1eb8b4f663e04a5685be2441dd502281cc","sha256:0b307cc08be91bc215dfad16e169a16c794359ecb00b469a19ddd446661b573b"],"state_sha256":"8a80355c862b6c2c503cbd1b47767d431766101c00ef038cf181ec57b46ed0b9"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"sDUCOa8gb44WshV7iysVtZYJoIbDlw2knqdqU2qNkFEnwnjWQmelfZx6Vac+9htuCJVXAOug2AkQ94z5cjWdDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-10T12:27:18.528929Z","bundle_sha256":"56baa86f7caaa1467e16d4668b835182d50c4e25d1813c5ecf8ee4dc3683f24c"}}