{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SLCQWKUBFTHJ6TVB3RTNVL7VKK","short_pith_number":"pith:SLCQWKUB","schema_version":"1.0","canonical_sha256":"92c50b2a812cce9f4ea1dc66daaff5528d3919d58c725cb5487d0b2a7b025a9c","source":{"kind":"arxiv","id":"2504.01871","version":1},"attestation_state":"computed","paper":{"title":"Interpreting Emergent Planning in Model-Free Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Adri\\`a Garriga-Alonso, David Krueger, Stephen Chung, Thomas Bush, Usman Anwar","submitted_at":"2025-04-02T16:24:23Z","abstract_excerpt":"We present the first mechanistic evidence that model-free reinforcement learning agents can learn to plan. This is achieved by applying a methodology based on concept-based interpretability to a model-free agent in Sokoban -- a commonly used benchmark for studying planning. Specifically, we demonstrate that DRC, a generic model-free agent introduced by Guez et al. (2019), uses learned concept representations to internally formulate plans that both predict the long-term effects of actions on the environment and influence action selection. Our methodology involves: (1) probing for planning-relev"},"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":"2504.01871","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-02T16:24:23Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8a15a86e59a16fc906c1dbe55bcd0dc2557574516d7033720b3e89b2282b3423","abstract_canon_sha256":"67fe64535e79cc4555adb34b1e2c481ea9220454cc247a90a7c9f43595506bf4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:43:31.032773Z","signature_b64":"hwP34YaUwPVBuqpVWyztk0KFsH4ctwVXdfJ8JYy5C+HkYSWge1c+QtkUDqpOzLNXRzL6URd9O2/GniacRLOsDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"92c50b2a812cce9f4ea1dc66daaff5528d3919d58c725cb5487d0b2a7b025a9c","last_reissued_at":"2026-07-05T10:43:31.032298Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:43:31.032298Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Interpreting Emergent Planning in Model-Free Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Adri\\`a Garriga-Alonso, David Krueger, Stephen Chung, Thomas Bush, Usman Anwar","submitted_at":"2025-04-02T16:24:23Z","abstract_excerpt":"We present the first mechanistic evidence that model-free reinforcement learning agents can learn to plan. This is achieved by applying a methodology based on concept-based interpretability to a model-free agent in Sokoban -- a commonly used benchmark for studying planning. Specifically, we demonstrate that DRC, a generic model-free agent introduced by Guez et al. (2019), uses learned concept representations to internally formulate plans that both predict the long-term effects of actions on the environment and influence action selection. Our methodology involves: (1) probing for planning-relev"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.01871","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/2504.01871/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":"2504.01871","created_at":"2026-07-05T10:43:31.032362+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.01871v1","created_at":"2026-07-05T10:43:31.032362+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.01871","created_at":"2026-07-05T10:43:31.032362+00:00"},{"alias_kind":"pith_short_12","alias_value":"SLCQWKUBFTHJ","created_at":"2026-07-05T10:43:31.032362+00:00"},{"alias_kind":"pith_short_16","alias_value":"SLCQWKUBFTHJ6TVB","created_at":"2026-07-05T10:43:31.032362+00:00"},{"alias_kind":"pith_short_8","alias_value":"SLCQWKUB","created_at":"2026-07-05T10:43:31.032362+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/SLCQWKUBFTHJ6TVB3RTNVL7VKK","json":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK.json","graph_json":"https://pith.science/api/pith-number/SLCQWKUBFTHJ6TVB3RTNVL7VKK/graph.json","events_json":"https://pith.science/api/pith-number/SLCQWKUBFTHJ6TVB3RTNVL7VKK/events.json","paper":"https://pith.science/paper/SLCQWKUB"},"agent_actions":{"view_html":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK","download_json":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK.json","view_paper":"https://pith.science/paper/SLCQWKUB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.01871&json=true","fetch_graph":"https://pith.science/api/pith-number/SLCQWKUBFTHJ6TVB3RTNVL7VKK/graph.json","fetch_events":"https://pith.science/api/pith-number/SLCQWKUBFTHJ6TVB3RTNVL7VKK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK/action/storage_attestation","attest_author":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK/action/author_attestation","sign_citation":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK/action/citation_signature","submit_replication":"https://pith.science/pith/SLCQWKUBFTHJ6TVB3RTNVL7VKK/action/replication_record"}},"created_at":"2026-07-05T10:43:31.032362+00:00","updated_at":"2026-07-05T10:43:31.032362+00:00"}