{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KZ26O2L2Z2BJB26ZRAG4DJ4Z3L","short_pith_number":"pith:KZ26O2L2","schema_version":"1.0","canonical_sha256":"5675e7697ace8290ebd9880dc1a799dac4f638ced8907c0507eecddc41f21106","source":{"kind":"arxiv","id":"2405.16012","version":3},"attestation_state":"computed","paper":{"title":"Pessimistic Backward Policy for GFlowNets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hyosoon Jang, Jinkyoo Park, Minsu Kim, Sungsoo Ahn, Yunhui Jang","submitted_at":"2024-05-25T02:30:46Z","abstract_excerpt":"This paper studies Generative Flow Networks (GFlowNets), which learn to sample objects proportionally to a given reward function through the trajectory of state transitions. In this work, we observe that GFlowNets tend to under-exploit the high-reward objects due to training on insufficient number of trajectories, which may lead to a large gap between the estimated flow and the (known) reward value. In response to this challenge, we propose a pessimistic backward policy for GFlowNets (PBP-GFN), which maximizes the observed flow to align closely with the true reward for the object. We extensive"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2405.16012","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-25T02:30:46Z","cross_cats_sorted":[],"title_canon_sha256":"26a6efc2146342efc57a5f35886204d89292ef5422c2cdbd416fd9d37406e06d","abstract_canon_sha256":"357a22316533092c38d1b3abfb0e8470d8ff5f6a4fc1b3bae4068f728846f0ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:27:22.303794Z","signature_b64":"NrfsLUf/Ko9MbjQe5NEnmGOB3+hR8wAVgGEqxCLSyi0YmGcw3/IXit49E8rNEnBAmx5kDdCld+GI807iUzaoDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5675e7697ace8290ebd9880dc1a799dac4f638ced8907c0507eecddc41f21106","last_reissued_at":"2026-07-05T09:27:22.303013Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:27:22.303013Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pessimistic Backward Policy for GFlowNets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hyosoon Jang, Jinkyoo Park, Minsu Kim, Sungsoo Ahn, Yunhui Jang","submitted_at":"2024-05-25T02:30:46Z","abstract_excerpt":"This paper studies Generative Flow Networks (GFlowNets), which learn to sample objects proportionally to a given reward function through the trajectory of state transitions. In this work, we observe that GFlowNets tend to under-exploit the high-reward objects due to training on insufficient number of trajectories, which may lead to a large gap between the estimated flow and the (known) reward value. In response to this challenge, we propose a pessimistic backward policy for GFlowNets (PBP-GFN), which maximizes the observed flow to align closely with the true reward for the object. We extensive"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.16012","kind":"arxiv","version":3},"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/2405.16012/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2405.16012","created_at":"2026-07-05T09:27:22.303128+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.16012v3","created_at":"2026-07-05T09:27:22.303128+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.16012","created_at":"2026-07-05T09:27:22.303128+00:00"},{"alias_kind":"pith_short_12","alias_value":"KZ26O2L2Z2BJ","created_at":"2026-07-05T09:27:22.303128+00:00"},{"alias_kind":"pith_short_16","alias_value":"KZ26O2L2Z2BJB26Z","created_at":"2026-07-05T09:27:22.303128+00:00"},{"alias_kind":"pith_short_8","alias_value":"KZ26O2L2","created_at":"2026-07-05T09:27:22.303128+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.20110","citing_title":"Beyond the Proxy: Trajectory-Distilled Guidance for Offline GFlowNet Training","ref_index":31,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L","json":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L.json","graph_json":"https://pith.science/api/pith-number/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/graph.json","events_json":"https://pith.science/api/pith-number/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/events.json","paper":"https://pith.science/paper/KZ26O2L2"},"agent_actions":{"view_html":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L","download_json":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L.json","view_paper":"https://pith.science/paper/KZ26O2L2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.16012&json=true","fetch_graph":"https://pith.science/api/pith-number/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/graph.json","fetch_events":"https://pith.science/api/pith-number/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/action/storage_attestation","attest_author":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/action/author_attestation","sign_citation":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/action/citation_signature","submit_replication":"https://pith.science/pith/KZ26O2L2Z2BJB26ZRAG4DJ4Z3L/action/replication_record"}},"created_at":"2026-07-05T09:27:22.303128+00:00","updated_at":"2026-07-05T09:27:22.303128+00:00"}