{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BTAQNZ6PJYC7KVEP4PV3ARZMT5","short_pith_number":"pith:BTAQNZ6P","schema_version":"1.0","canonical_sha256":"0cc106e7cf4e05f5548fe3ebb0472c9f75c24ce9b795ff1fcc3e9c11f2e5e8a8","source":{"kind":"arxiv","id":"2509.10423","version":1},"attestation_state":"computed","paper":{"title":"Mutual Information Tracks Policy Coherence in Reinforcement Learning","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG","cs.RO"],"primary_cat":"cs.AI","authors_text":"Amirhossein Nazeri, Cameron Reid, Wael Hafez","submitted_at":"2025-09-12T17:24:20Z","abstract_excerpt":"Reinforcement Learning (RL) agents deployed in real-world environments face degradation from sensor faults, actuator wear, and environmental shifts, yet lack intrinsic mechanisms to detect and diagnose these failures. We present an information-theoretic framework that reveals both the fundamental dynamics of RL and provides practical methods for diagnosing deployment-time anomalies. Through analysis of state-action mutual information patterns in a robotic control task, we first demonstrate that successful learning exhibits characteristic information signatures: mutual information between state"},"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":"2509.10423","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.AI","submitted_at":"2025-09-12T17:24:20Z","cross_cats_sorted":["cs.LG","cs.RO"],"title_canon_sha256":"4d46e67aaa34185a86726629562c8ccb3f95e513c895422d4c1b67ef4a9e4f0a","abstract_canon_sha256":"ea3ab6f0ac2917d64fce19db830dc67a6b06db5bdcea2b81b34ee7fb430780c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:11:09.511814Z","signature_b64":"FSrg2LiKbFKUOFPOanI/oTiMEqM/fVeRsMaSZ77qfCvW1Kdmyk25zxqaPvbb9/c+s+X2vBVDa6/I1/p1AZ70Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0cc106e7cf4e05f5548fe3ebb0472c9f75c24ce9b795ff1fcc3e9c11f2e5e8a8","last_reissued_at":"2026-07-05T12:11:09.511293Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:11:09.511293Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mutual Information Tracks Policy Coherence in Reinforcement Learning","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG","cs.RO"],"primary_cat":"cs.AI","authors_text":"Amirhossein Nazeri, Cameron Reid, Wael Hafez","submitted_at":"2025-09-12T17:24:20Z","abstract_excerpt":"Reinforcement Learning (RL) agents deployed in real-world environments face degradation from sensor faults, actuator wear, and environmental shifts, yet lack intrinsic mechanisms to detect and diagnose these failures. We present an information-theoretic framework that reveals both the fundamental dynamics of RL and provides practical methods for diagnosing deployment-time anomalies. Through analysis of state-action mutual information patterns in a robotic control task, we first demonstrate that successful learning exhibits characteristic information signatures: mutual information between state"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.10423","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/2509.10423/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":"2509.10423","created_at":"2026-07-05T12:11:09.511361+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.10423v1","created_at":"2026-07-05T12:11:09.511361+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.10423","created_at":"2026-07-05T12:11:09.511361+00:00"},{"alias_kind":"pith_short_12","alias_value":"BTAQNZ6PJYC7","created_at":"2026-07-05T12:11:09.511361+00:00"},{"alias_kind":"pith_short_16","alias_value":"BTAQNZ6PJYC7KVEP","created_at":"2026-07-05T12:11:09.511361+00:00"},{"alias_kind":"pith_short_8","alias_value":"BTAQNZ6P","created_at":"2026-07-05T12:11:09.511361+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26627","citing_title":"Breaking the Epistemic Trap: Active Perception Under Compound Uncertainty","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26627","citing_title":"Breaking the Epistemic Trap: Active Perception Under Compound Uncertainty","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2603.01283","citing_title":"The Informational Cost of Agency: A Bounded Measure of Interaction Efficiency for Deployed Reinforcement Learning","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2603.01283","citing_title":"The Informational Cost of Agency: A Bounded Measure of Interaction Efficiency for Deployed Reinforcement Learning","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5","json":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5.json","graph_json":"https://pith.science/api/pith-number/BTAQNZ6PJYC7KVEP4PV3ARZMT5/graph.json","events_json":"https://pith.science/api/pith-number/BTAQNZ6PJYC7KVEP4PV3ARZMT5/events.json","paper":"https://pith.science/paper/BTAQNZ6P"},"agent_actions":{"view_html":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5","download_json":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5.json","view_paper":"https://pith.science/paper/BTAQNZ6P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.10423&json=true","fetch_graph":"https://pith.science/api/pith-number/BTAQNZ6PJYC7KVEP4PV3ARZMT5/graph.json","fetch_events":"https://pith.science/api/pith-number/BTAQNZ6PJYC7KVEP4PV3ARZMT5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5/action/storage_attestation","attest_author":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5/action/author_attestation","sign_citation":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5/action/citation_signature","submit_replication":"https://pith.science/pith/BTAQNZ6PJYC7KVEP4PV3ARZMT5/action/replication_record"}},"created_at":"2026-07-05T12:11:09.511361+00:00","updated_at":"2026-07-05T12:11:09.511361+00:00"}