{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EQ6MBZGBOY2G3L66RRSW4L7BGU","short_pith_number":"pith:EQ6MBZGB","schema_version":"1.0","canonical_sha256":"243cc0e4c176346dafde8c656e2fe13532bb14690337a680085b01e880252983","source":{"kind":"arxiv","id":"2605.30854","version":1},"attestation_state":"computed","paper":{"title":"Safe Equilibrium Policy Optimization for Strategic Agent Policies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.MA","authors_text":"Amit Dhanda, Karthika Arumugam, Kiran Kumar Manku","submitted_at":"2026-05-29T05:20:32Z","abstract_excerpt":"Language models fine-tuned with reinforcement learning typically optimize for task reward, ignoring multi-agent strategic structure. Because these agents condition on natural language game-state descriptions and emit actions through free-form generation, strategic failure modes -- exploiting weaker opponents, coordinating on harmful equilibria, and externalizing costs are inseparable from the language interface itself. We propose Safe Equilibrium Policy Optimization (\\sepo{}), a training objective that augments expected payoff with explicit penalties for exploitability, collusion risk, and ext"},"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":"2605.30854","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.MA","submitted_at":"2026-05-29T05:20:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"71c2f88dab5f28d9273d7a66cd3a43b4e7c0fcae94b1352510c7517fbcbfda34","abstract_canon_sha256":"e330dfef37c8f01bb5511ef0c0265018b818e6bc24935b34057f791b0ebbef58"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:03:21.146781Z","signature_b64":"mWpbrwkD6a+8eouGGz7GUt8wZukrIj7Pl6uvj8tEqGA40ACoOUhIdcTlq0qksmmHcrTyBqiMTQNu/uf98gSKCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"243cc0e4c176346dafde8c656e2fe13532bb14690337a680085b01e880252983","last_reissued_at":"2026-06-01T01:03:21.146043Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:03:21.146043Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Safe Equilibrium Policy Optimization for Strategic Agent Policies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.MA","authors_text":"Amit Dhanda, Karthika Arumugam, Kiran Kumar Manku","submitted_at":"2026-05-29T05:20:32Z","abstract_excerpt":"Language models fine-tuned with reinforcement learning typically optimize for task reward, ignoring multi-agent strategic structure. Because these agents condition on natural language game-state descriptions and emit actions through free-form generation, strategic failure modes -- exploiting weaker opponents, coordinating on harmful equilibria, and externalizing costs are inseparable from the language interface itself. We propose Safe Equilibrium Policy Optimization (\\sepo{}), a training objective that augments expected payoff with explicit penalties for exploitability, collusion risk, and ext"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.30854","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/2605.30854/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":"2605.30854","created_at":"2026-06-01T01:03:21.146155+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.30854v1","created_at":"2026-06-01T01:03:21.146155+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.30854","created_at":"2026-06-01T01:03:21.146155+00:00"},{"alias_kind":"pith_short_12","alias_value":"EQ6MBZGBOY2G","created_at":"2026-06-01T01:03:21.146155+00:00"},{"alias_kind":"pith_short_16","alias_value":"EQ6MBZGBOY2G3L66","created_at":"2026-06-01T01:03:21.146155+00:00"},{"alias_kind":"pith_short_8","alias_value":"EQ6MBZGB","created_at":"2026-06-01T01:03:21.146155+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/EQ6MBZGBOY2G3L66RRSW4L7BGU","json":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU.json","graph_json":"https://pith.science/api/pith-number/EQ6MBZGBOY2G3L66RRSW4L7BGU/graph.json","events_json":"https://pith.science/api/pith-number/EQ6MBZGBOY2G3L66RRSW4L7BGU/events.json","paper":"https://pith.science/paper/EQ6MBZGB"},"agent_actions":{"view_html":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU","download_json":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU.json","view_paper":"https://pith.science/paper/EQ6MBZGB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.30854&json=true","fetch_graph":"https://pith.science/api/pith-number/EQ6MBZGBOY2G3L66RRSW4L7BGU/graph.json","fetch_events":"https://pith.science/api/pith-number/EQ6MBZGBOY2G3L66RRSW4L7BGU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU/action/storage_attestation","attest_author":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU/action/author_attestation","sign_citation":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU/action/citation_signature","submit_replication":"https://pith.science/pith/EQ6MBZGBOY2G3L66RRSW4L7BGU/action/replication_record"}},"created_at":"2026-06-01T01:03:21.146155+00:00","updated_at":"2026-06-01T01:03:21.146155+00:00"}