{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:NCRPO36CDC7IUYAO7WC44X4PYX","short_pith_number":"pith:NCRPO36C","schema_version":"1.0","canonical_sha256":"68a2f76fc218be8a600efd85ce5f8fc5c97fcd6530427ceafa32fc4763d098ee","source":{"kind":"arxiv","id":"2102.09756","version":2},"attestation_state":"computed","paper":{"title":"TacticZero: Learning to Prove Theorems from Scratch with Deep Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LO"],"primary_cat":"cs.LG","authors_text":"Amir Dezfouli, Christian Walder, Michael Norrish, Minchao Wu","submitted_at":"2021-02-19T06:08:39Z","abstract_excerpt":"We propose a novel approach to interactive theorem-proving (ITP) using deep reinforcement learning. The proposed framework is able to learn proof search strategies as well as tactic and arguments prediction in an end-to-end manner. We formulate the process of ITP as a Markov decision process (MDP) in which each state represents a set of potential derivation paths. This structure allows us to introduce a novel backtracking mechanism which enables the agent to efficiently discard (predicted) dead-end derivations and restart from promising alternatives. We implement the framework in the HOL4 theo"},"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":"2102.09756","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-02-19T06:08:39Z","cross_cats_sorted":["cs.AI","cs.LO"],"title_canon_sha256":"7f6953b0a1ffb0f0949cd6804daae78e0b3eb48fcbdd3981a9c0c381392c7c0c","abstract_canon_sha256":"2eabe69f7b676f074b0418853d0ce7f69af3dbaa6e0d6600149049116d8dca86"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:50:09.328746Z","signature_b64":"cf49EhqZyFnkL+pNONqAW6xf/o7NlpS8B0dc6ZQbNLKgAiU015peQQFg3OTED5dLe9QFg9fzL/8fJFs1X2StBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"68a2f76fc218be8a600efd85ce5f8fc5c97fcd6530427ceafa32fc4763d098ee","last_reissued_at":"2026-07-05T02:50:09.328355Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:50:09.328355Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TacticZero: Learning to Prove Theorems from Scratch with Deep Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LO"],"primary_cat":"cs.LG","authors_text":"Amir Dezfouli, Christian Walder, Michael Norrish, Minchao Wu","submitted_at":"2021-02-19T06:08:39Z","abstract_excerpt":"We propose a novel approach to interactive theorem-proving (ITP) using deep reinforcement learning. The proposed framework is able to learn proof search strategies as well as tactic and arguments prediction in an end-to-end manner. We formulate the process of ITP as a Markov decision process (MDP) in which each state represents a set of potential derivation paths. This structure allows us to introduce a novel backtracking mechanism which enables the agent to efficiently discard (predicted) dead-end derivations and restart from promising alternatives. We implement the framework in the HOL4 theo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2102.09756","kind":"arxiv","version":2},"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/2102.09756/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":"2102.09756","created_at":"2026-07-05T02:50:09.328414+00:00"},{"alias_kind":"arxiv_version","alias_value":"2102.09756v2","created_at":"2026-07-05T02:50:09.328414+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2102.09756","created_at":"2026-07-05T02:50:09.328414+00:00"},{"alias_kind":"pith_short_12","alias_value":"NCRPO36CDC7I","created_at":"2026-07-05T02:50:09.328414+00:00"},{"alias_kind":"pith_short_16","alias_value":"NCRPO36CDC7IUYAO","created_at":"2026-07-05T02:50:09.328414+00:00"},{"alias_kind":"pith_short_8","alias_value":"NCRPO36C","created_at":"2026-07-05T02:50:09.328414+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.14063","citing_title":"Rango: Adaptive Retrieval-Augmented Proving for Automated Software Verification","ref_index":87,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX","json":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX.json","graph_json":"https://pith.science/api/pith-number/NCRPO36CDC7IUYAO7WC44X4PYX/graph.json","events_json":"https://pith.science/api/pith-number/NCRPO36CDC7IUYAO7WC44X4PYX/events.json","paper":"https://pith.science/paper/NCRPO36C"},"agent_actions":{"view_html":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX","download_json":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX.json","view_paper":"https://pith.science/paper/NCRPO36C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2102.09756&json=true","fetch_graph":"https://pith.science/api/pith-number/NCRPO36CDC7IUYAO7WC44X4PYX/graph.json","fetch_events":"https://pith.science/api/pith-number/NCRPO36CDC7IUYAO7WC44X4PYX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX/action/storage_attestation","attest_author":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX/action/author_attestation","sign_citation":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX/action/citation_signature","submit_replication":"https://pith.science/pith/NCRPO36CDC7IUYAO7WC44X4PYX/action/replication_record"}},"created_at":"2026-07-05T02:50:09.328414+00:00","updated_at":"2026-07-05T02:50:09.328414+00:00"}