{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PP2VIP7AY6EHVHIOVQR2STQYLG","short_pith_number":"pith:PP2VIP7A","schema_version":"1.0","canonical_sha256":"7bf5543fe0c7887a9d0eac23a94e1859a22eb063dc88991a8b2aeeb62303efd0","source":{"kind":"arxiv","id":"2606.24994","version":1},"attestation_state":"computed","paper":{"title":"ExTra: Exploratory Trajectory Optimization for Language Model Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bryan Kian Hsiang Low, Daniel Dahlmeier, Junxiang Jia, See-kiong Ng, Wenyang Hu, Zhen Shu","submitted_at":"2026-06-23T15:51:39Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) for language-model reasoning can fail at both extremes of task difficulty: easy prompts often produce all-correct, low-diversity rollout groups with little gradient signal, while hard prompts can produce all-incorrect groups with no positive reward. We introduce ExTra (Exploratory Trajectory Optimization), a GRPO-compatible framework that extracts exploration signals from the model's own rollouts. ExTra combines two mechanisms: (i) a novelty reward that adds embedding-based diversity bonuses after GRPO normalization, rewarding diverse corre"},"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":"2606.24994","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-23T15:51:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e43fbf5c7df358b34785b0dee35f5a6b787022449d592849908f2eb2b3b5c5f3","abstract_canon_sha256":"e6d2cbbb2d0ce1d8b47973c09eb1a7695868b792eeff40067d62387f606ac0ef"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T00:17:47.769441Z","signature_b64":"Azsa44KPaKIgrsIKiskjFYQqPeHZYftcQ/JWBr6Kfy6WzbKArfGpKlXfy83u9gjwVpD7l1/M1miNpUHDoD1tBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7bf5543fe0c7887a9d0eac23a94e1859a22eb063dc88991a8b2aeeb62303efd0","last_reissued_at":"2026-06-25T00:17:47.769039Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T00:17:47.769039Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ExTra: Exploratory Trajectory Optimization for Language Model Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bryan Kian Hsiang Low, Daniel Dahlmeier, Junxiang Jia, See-kiong Ng, Wenyang Hu, Zhen Shu","submitted_at":"2026-06-23T15:51:39Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) for language-model reasoning can fail at both extremes of task difficulty: easy prompts often produce all-correct, low-diversity rollout groups with little gradient signal, while hard prompts can produce all-incorrect groups with no positive reward. We introduce ExTra (Exploratory Trajectory Optimization), a GRPO-compatible framework that extracts exploration signals from the model's own rollouts. ExTra combines two mechanisms: (i) a novelty reward that adds embedding-based diversity bonuses after GRPO normalization, rewarding diverse corre"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.24994","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/2606.24994/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":"2606.24994","created_at":"2026-06-25T00:17:47.769103+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.24994v1","created_at":"2026-06-25T00:17:47.769103+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.24994","created_at":"2026-06-25T00:17:47.769103+00:00"},{"alias_kind":"pith_short_12","alias_value":"PP2VIP7AY6EH","created_at":"2026-06-25T00:17:47.769103+00:00"},{"alias_kind":"pith_short_16","alias_value":"PP2VIP7AY6EHVHIO","created_at":"2026-06-25T00:17:47.769103+00:00"},{"alias_kind":"pith_short_8","alias_value":"PP2VIP7A","created_at":"2026-06-25T00:17:47.769103+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/PP2VIP7AY6EHVHIOVQR2STQYLG","json":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG.json","graph_json":"https://pith.science/api/pith-number/PP2VIP7AY6EHVHIOVQR2STQYLG/graph.json","events_json":"https://pith.science/api/pith-number/PP2VIP7AY6EHVHIOVQR2STQYLG/events.json","paper":"https://pith.science/paper/PP2VIP7A"},"agent_actions":{"view_html":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG","download_json":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG.json","view_paper":"https://pith.science/paper/PP2VIP7A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.24994&json=true","fetch_graph":"https://pith.science/api/pith-number/PP2VIP7AY6EHVHIOVQR2STQYLG/graph.json","fetch_events":"https://pith.science/api/pith-number/PP2VIP7AY6EHVHIOVQR2STQYLG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG/action/storage_attestation","attest_author":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG/action/author_attestation","sign_citation":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG/action/citation_signature","submit_replication":"https://pith.science/pith/PP2VIP7AY6EHVHIOVQR2STQYLG/action/replication_record"}},"created_at":"2026-06-25T00:17:47.769103+00:00","updated_at":"2026-06-25T00:17:47.769103+00:00"}