{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:EZCTUFPZR6A5URHZPN7NIZDUOK","short_pith_number":"pith:EZCTUFPZ","canonical_record":{"source":{"id":"2504.20887","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-29T16:04:16Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dd06710345a7dc9aed54004132066a4331c195d1b1c34b446a4312ad7ef9eb5f","abstract_canon_sha256":"54567a607c572a4ebdfb5479f37985a259434dabcd46d269305b0e49cf2ad623"},"schema_version":"1.0"},"canonical_sha256":"26453a15f98f81da44f97b7ed4647472962533feefb6255cf95f077cacb9a75d","source":{"kind":"arxiv","id":"2504.20887","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.20887","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"arxiv_version","alias_value":"2504.20887v2","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.20887","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"pith_short_12","alias_value":"EZCTUFPZR6A5","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"pith_short_16","alias_value":"EZCTUFPZR6A5URHZ","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"pith_short_8","alias_value":"EZCTUFPZ","created_at":"2026-07-05T11:39:53Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:EZCTUFPZR6A5URHZPN7NIZDUOK","target":"record","payload":{"canonical_record":{"source":{"id":"2504.20887","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-29T16:04:16Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dd06710345a7dc9aed54004132066a4331c195d1b1c34b446a4312ad7ef9eb5f","abstract_canon_sha256":"54567a607c572a4ebdfb5479f37985a259434dabcd46d269305b0e49cf2ad623"},"schema_version":"1.0"},"canonical_sha256":"26453a15f98f81da44f97b7ed4647472962533feefb6255cf95f077cacb9a75d","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:39:53.451472Z","signature_b64":"Sw7lYzi31ujuAUUU0UTcDUrK1ixv2KR23n+Oqbi9l5CQ1KzIMFkqcKXTltcputrID3VsEQFdsy1402tJd0w0Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"26453a15f98f81da44f97b7ed4647472962533feefb6255cf95f077cacb9a75d","last_reissued_at":"2026-07-05T11:39:53.450920Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:39:53.450920Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.20887","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":"2026-07-05T11:39:53Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xnOVPSwnt+15P3Dr4rn8EhR9phe/+SRh/iyvGQN340joQQeajkFtAfA6jLVT/IsGNvjYMJboHcFD+axA6bR4Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T18:28:39.926131Z"},"content_sha256":"6a5ce880cfa27f8bd453d3de4ee7d7283ca859e4ac6a0a41b5ba1c632b1ce14d","schema_version":"1.0","event_id":"sha256:6a5ce880cfa27f8bd453d3de4ee7d7283ca859e4ac6a0a41b5ba1c632b1ce14d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:EZCTUFPZR6A5URHZPN7NIZDUOK","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Return Capping: Sample-Efficient CVaR Policy Gradient Optimisation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bruno Lacerda, Clarissa Costen, Harry Mead, Nick Hawes","submitted_at":"2025-04-29T16:04:16Z","abstract_excerpt":"When optimising for conditional value at risk (CVaR) using policy gradients (PG), current methods rely on discarding a large proportion of trajectories, resulting in poor sample efficiency. We propose a reformulation of the CVaR optimisation problem by capping the total return of trajectories used in training, rather than simply discarding them, and show that this is equivalent to the original problem if the cap is set appropriately. We show, with empirical results in an number of environments, that this reformulation of the problem results in consistently improved performance compared to base"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.20887","kind":"arxiv","version":2},"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/2504.20887/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:39:53Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xcCmz/1Z32B+uEnY/ROMl3iOgn3JTVge0k6SDKEpBeBjWgC/vXDrpwmrIhmSf8LcQQvDBkEHPMtu67PAEEWRAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T18:28:39.926763Z"},"content_sha256":"fbe694b0a35b67bfa39c2f7ece29330a89cef8c5fc03376ee7c4f4ee59e8359c","schema_version":"1.0","event_id":"sha256:fbe694b0a35b67bfa39c2f7ece29330a89cef8c5fc03376ee7c4f4ee59e8359c"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/EZCTUFPZR6A5URHZPN7NIZDUOK/bundle.json","state_url":"https://pith.science/pith/EZCTUFPZR6A5URHZPN7NIZDUOK/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/EZCTUFPZR6A5URHZPN7NIZDUOK/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":"2026-08-10T18:28:39Z","links":{"resolver":"https://pith.science/pith/EZCTUFPZR6A5URHZPN7NIZDUOK","bundle":"https://pith.science/pith/EZCTUFPZR6A5URHZPN7NIZDUOK/bundle.json","state":"https://pith.science/pith/EZCTUFPZR6A5URHZPN7NIZDUOK/state.json","well_known_bundle":"https://pith.science/.well-known/pith/EZCTUFPZR6A5URHZPN7NIZDUOK/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:EZCTUFPZR6A5URHZPN7NIZDUOK","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":"54567a607c572a4ebdfb5479f37985a259434dabcd46d269305b0e49cf2ad623","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-29T16:04:16Z","title_canon_sha256":"dd06710345a7dc9aed54004132066a4331c195d1b1c34b446a4312ad7ef9eb5f"},"schema_version":"1.0","source":{"id":"2504.20887","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.20887","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"arxiv_version","alias_value":"2504.20887v2","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.20887","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"pith_short_12","alias_value":"EZCTUFPZR6A5","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"pith_short_16","alias_value":"EZCTUFPZR6A5URHZ","created_at":"2026-07-05T11:39:53Z"},{"alias_kind":"pith_short_8","alias_value":"EZCTUFPZ","created_at":"2026-07-05T11:39:53Z"}],"graph_snapshots":[{"event_id":"sha256:fbe694b0a35b67bfa39c2f7ece29330a89cef8c5fc03376ee7c4f4ee59e8359c","target":"graph","created_at":"2026-07-05T11:39:53Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2504.20887/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"When optimising for conditional value at risk (CVaR) using policy gradients (PG), current methods rely on discarding a large proportion of trajectories, resulting in poor sample efficiency. We propose a reformulation of the CVaR optimisation problem by capping the total return of trajectories used in training, rather than simply discarding them, and show that this is equivalent to the original problem if the cap is set appropriately. We show, with empirical results in an number of environments, that this reformulation of the problem results in consistently improved performance compared to base","authors_text":"Bruno Lacerda, Clarissa Costen, Harry Mead, Nick Hawes","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-29T16:04:16Z","title":"Return Capping: Sample-Efficient CVaR Policy Gradient Optimisation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.20887","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:6a5ce880cfa27f8bd453d3de4ee7d7283ca859e4ac6a0a41b5ba1c632b1ce14d","target":"record","created_at":"2026-07-05T11:39:53Z","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":"54567a607c572a4ebdfb5479f37985a259434dabcd46d269305b0e49cf2ad623","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-29T16:04:16Z","title_canon_sha256":"dd06710345a7dc9aed54004132066a4331c195d1b1c34b446a4312ad7ef9eb5f"},"schema_version":"1.0","source":{"id":"2504.20887","kind":"arxiv","version":2}},"canonical_sha256":"26453a15f98f81da44f97b7ed4647472962533feefb6255cf95f077cacb9a75d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"26453a15f98f81da44f97b7ed4647472962533feefb6255cf95f077cacb9a75d","first_computed_at":"2026-07-05T11:39:53.450920Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:39:53.450920Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Sw7lYzi31ujuAUUU0UTcDUrK1ixv2KR23n+Oqbi9l5CQ1KzIMFkqcKXTltcputrID3VsEQFdsy1402tJd0w0Cg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:39:53.451472Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.20887","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6a5ce880cfa27f8bd453d3de4ee7d7283ca859e4ac6a0a41b5ba1c632b1ce14d","sha256:fbe694b0a35b67bfa39c2f7ece29330a89cef8c5fc03376ee7c4f4ee59e8359c"],"state_sha256":"3c270c6c53105ac273fd47ddbf06b7d3a04fa4d02eaf22589732719b7422d26b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JnKGmkC+Grf/bdAVkk3Wn/0WShxEwXv/xbBwIwylgeQYDIav5fkORsAiaow61mP22TaZZ/skCTohXZgMGF70Cg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T18:28:39.932693Z","bundle_sha256":"7eb258a442aa8ac3eab48153e13119fe284f562b3987d4737902bb6dcbb94c0d"}}