{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DUXT2OPL72AJ5EBL46VPYEK7SA","short_pith_number":"pith:DUXT2OPL","schema_version":"1.0","canonical_sha256":"1d2f3d39ebfe809e902be7aafc115f902376e640100553161038b9e9682ccf59","source":{"kind":"arxiv","id":"2508.18884","version":1},"attestation_state":"computed","paper":{"title":"HAEPO: History-Aggregated Exploratory Policy Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Alakh Sharma, Dhruv Kumar, Gaurish Trivedi, Jagat Sesh Challa, Kartikey Singh Bhandari, Pratik Narang","submitted_at":"2025-08-26T09:59:44Z","abstract_excerpt":"Exploration is essential in modern learning, from reinforcement learning environments with small neural policies to large language models (LLMs). Existing work, such as DPO, leverages full sequence log-likelihoods to capture an entire trajectory of the model's decisions, while methods like GRPO aggregate per-token ratios into a trajectory-level update. However, both often limit exploration on long-horizon tasks. We introduce History-Aggregated Exploratory Policy Optimization (HAEPO), a history-aware exploratory loss to combat these shortcomings. HAEPO compresses each trajectory into the sum of"},"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":"2508.18884","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-26T09:59:44Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d803f3281c303e3afa086b8e61e0f5bd53b136466cf92721fb587e5233297849","abstract_canon_sha256":"c1e91c12841991675f5031397acde6993e72b8b3823e498e0742058b7c74b7ca"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:59:32.988414Z","signature_b64":"ETi9ERqp8JV+gMPnA3XTxASb8L77IbI8sw7+tCmoBa80wqw5KsNMX5JN6EOdHwbsg0dqu8AfaTAN2qCGhvEdCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1d2f3d39ebfe809e902be7aafc115f902376e640100553161038b9e9682ccf59","last_reissued_at":"2026-07-05T11:59:32.987937Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:59:32.987937Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HAEPO: History-Aggregated Exploratory Policy Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Alakh Sharma, Dhruv Kumar, Gaurish Trivedi, Jagat Sesh Challa, Kartikey Singh Bhandari, Pratik Narang","submitted_at":"2025-08-26T09:59:44Z","abstract_excerpt":"Exploration is essential in modern learning, from reinforcement learning environments with small neural policies to large language models (LLMs). Existing work, such as DPO, leverages full sequence log-likelihoods to capture an entire trajectory of the model's decisions, while methods like GRPO aggregate per-token ratios into a trajectory-level update. However, both often limit exploration on long-horizon tasks. We introduce History-Aggregated Exploratory Policy Optimization (HAEPO), a history-aware exploratory loss to combat these shortcomings. HAEPO compresses each trajectory into the sum of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.18884","kind":"arxiv","version":1},"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/2508.18884/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":"2508.18884","created_at":"2026-07-05T11:59:32.988002+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.18884v1","created_at":"2026-07-05T11:59:32.988002+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.18884","created_at":"2026-07-05T11:59:32.988002+00:00"},{"alias_kind":"pith_short_12","alias_value":"DUXT2OPL72AJ","created_at":"2026-07-05T11:59:32.988002+00:00"},{"alias_kind":"pith_short_16","alias_value":"DUXT2OPL72AJ5EBL","created_at":"2026-07-05T11:59:32.988002+00:00"},{"alias_kind":"pith_short_8","alias_value":"DUXT2OPL","created_at":"2026-07-05T11:59:32.988002+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA","json":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA.json","graph_json":"https://pith.science/api/pith-number/DUXT2OPL72AJ5EBL46VPYEK7SA/graph.json","events_json":"https://pith.science/api/pith-number/DUXT2OPL72AJ5EBL46VPYEK7SA/events.json","paper":"https://pith.science/paper/DUXT2OPL"},"agent_actions":{"view_html":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA","download_json":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA.json","view_paper":"https://pith.science/paper/DUXT2OPL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.18884&json=true","fetch_graph":"https://pith.science/api/pith-number/DUXT2OPL72AJ5EBL46VPYEK7SA/graph.json","fetch_events":"https://pith.science/api/pith-number/DUXT2OPL72AJ5EBL46VPYEK7SA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA/action/storage_attestation","attest_author":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA/action/author_attestation","sign_citation":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA/action/citation_signature","submit_replication":"https://pith.science/pith/DUXT2OPL72AJ5EBL46VPYEK7SA/action/replication_record"}},"created_at":"2026-07-05T11:59:32.988002+00:00","updated_at":"2026-07-05T11:59:32.988002+00:00"}